Mini Project - 1 A Real Time Object Detection

Contents
1. Introduction....................................................................................................................................2

1.1 Real Time Object Detection.....................................................................................................2

2. Aim and Objective..........................................................................................................................3

2.1 Aims.........................................................................................................................................3

2.2 Objective...................................................................................................................................4

3. Literature Survey...........................................................................................................................5

Paper l..................................................................................................................................................5

Paper ll.................................................................................................................................................6

Paper lll................................................................................................................................................7

Comparative Analysis..........................................................................................................................8

4. Existing System...............................................................................................................................9

5. Problem Statement.......................................................................................................................10

6. Scope..............................................................................................................................................11

7. Proposed System...........................................................................................................................13

7.1 System Architecture................................................................................................................14

7.2 ALGORITHM........................................................................................................................15

8. Planning.........................................................................................................................................21

9. Designing details...........................................................................................................................22

9.1 Coding....................................................................................................................................22

Screenshots........................................................................................................................................26

10. Conclusion.................................................................................................................................27

11. Requirements............................................................................................................................28

11.1 Hardware Requirements.........................................................................................................28

11.2 Software Requirements...........................................................................................................28

12. References..................................................................................................................................29

SSJCET, Asangaon 1 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

1. Introduction
1.1 Real Time Object Detection
Object detection is a crucial task in computer vision and has numerous applications,
ranging from surveillance systems to autonomous vehicles. The objective of this mini project
is to develop a system that can accurately detect and localize objects in images using
pretrained deep learning models.
In this report, we will discuss the methodology and results of implementing an object
detection system. The project involves leveraging state-of-the-art deep learning models,
specifically, the Single Shot Multi Box Detector (SSD) model, and applying it to a set of test
images to detect objects of interest.
The primary goal of this project is to gain practical experience in deploying pre-
trained models, understanding the underlying concepts of object detection, and evaluating the
performance of the system. Throughout the report, we will highlight the steps involved in the
implementation process, challenges faced, and solutions derived.
We will also discuss the dataset used for training and evaluation, the metrics utilized
to measure the accuracy of the object detection system, and any limitations encountered in the
project.
Finally, we will analyse the results obtained by the system, compare them with the
ground truth annotations, and provide recommendations for future enhancements or
optimizations. The report will conclude with a summary of the key findings and an assessment
of the overall success of the project.
By the end of this report, readers will have a comprehensive understanding of the
object detection process, the practical considerations involved in implementing a real-world
system, and the potential applications and limitations of the developed solution.

SSJCET, Asangaon 2 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

2. Aim and Objective

2.1 Aims
The aim of object detection is to accurately identify and locate specific objects within
an image or video. This task goes beyond simple image classification, as it involves not only
determining what objects are present but also their precise locations in the image.

The primary goal of object detection is to provide an automated and efficient solution
that can identify and localize objects in various real-world scenarios. This technology has
numerous applications, including but not limited to:

• Surveillance and security systems: Object detection can be used to monitor and
identify potential threats or suspicious activities in surveillance footage.

• Autonomous vehicles: Object detection is crucial for autonomous vehicles to detect

and track objects such as pedestrians, vehicles, and traffic signs to ensure safe
navigation.

• Retail and inventory management: By detecting objects, businesses can track

inventory, monitor stock levels, and even optimize product placement in retail
environments.

• Agriculture: Object detection can be used to monitor crop health, identify weed
growth, and detect pests or diseases in agricultural fields.

• Medical imaging: Object detection techniques can assist in identifying specific

anatomical structures or abnormalities in medical images, aiding in diagnosis and
treatment planning.

The aim of object detection is to develop accurate and efficient algorithms that can
operate in real-time, handle complex scenes, and generalize well to a wide range of objects
and environments.

SSJCET, Asangaon 3 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

2.2 Objective
The objectives of object detection can be summarized as follows:

 Accurate detection: The primary objective of object detection is to accurately identify

and localize objects of interest in images or videos. The system should be able to detect
objects with a high level of accuracy, minimizing false positives and false negatives.
 Localization: Object detection aims to provide precise location information for
detected objects. This includes drawing bounding boxes around the objects to indicate their
position and size within the image.
 Real-time performance: Object detection systems are often deployed in real-time
scenarios, such as surveillance or autonomous vehicles. Therefore, a key objective is to
achieve fast and efficient detection, enabling the system to process and analyze images or
video frames in real-time without significant delays.
 Robustness: Object detection algorithms should be able to handle a wide variety of
objects, diverse backgrounds, lighting conditions, and viewpoints. The system should be
robust enough to accurately detect objects in different environments and under different
circumstances.
 Scalability: Object detection systems should be able to handle large datasets and scale
well as the number of objects and complexity of the scenes increase. This includes efficient
memory usage, parallel processing capabilities, and the ability to handle a large number of
object classes.
 Generalization: Object detection algorithms should have the ability to generalize well
to unseen objects or object categories. The system should not be overly dependent on specific
training examples but should be able to apply learned knowledge to detect similar objects in
new situations.
 Integration and deployment: Object detection systems should be easily integrated into
existing workflows and applications. This may involve packaging the system as a software
library or API that can be easily integrated into other applications or systems.

SSJCET, Asangaon 4 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

3. Literature Survey

Paper l

Paper Name: You Only Look Once: Unified, Real-Time Object Detection

Author Name: Joseph Redmon, Santosh Divvala, Ross Girshick, Ali Farhadi

EXPLANATION:

Object detection traditionally relied on slow and computationally expensive methods

that used region proposal networks (e.g., R-CNN, Fast R-CNN). These methods divided
object detection into two stages: generating proposals for possible object locations and then
classifying those regions.

The authors introduced YOLO to streamline this process. Instead of treating object detection
as a two-stage problem, YOLO framed it as a single regression problem, enabling faster
detection by predicting bounding boxes and class labels in one pass.

The YOLO report introduced a paradigm shift in object detection by prioritizing speed and
simplicity. YOLO’s unified approach demonstrated that object detection models could
achieve real-time performance without the need for complex pipelines, making it one of the
most influential works in the field of computer vision.

SSJCET, Asangaon 5 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

Paper ll

PAPER NAME: R-CNN (Region-based Convolutional Neural Networks)

Author name: Ross B. Girshick, Jeff Donahue, Trevor Darrell, Jitendra Malik

Explanation:

Before R-CNN, traditional object detection models relied on hand-crafted features

(such as HOG or SIFT) and shallow classifiers (like SVMs), which were often
computationally expensive and lacked accuracy in complex scenes. With the rise of CNNs,
there was a growing interest in using deep learning to automatically learn powerful image
features, but applying CNNs directly to the entire image was inefficient for object detection.

The main challenge in object detection was efficiently localizing and classifying multiple
objects within an image. R-CNN solved this by focusing on region-based object detection,
combining region proposals and CNN feature extraction.

R-CNN was a landmark model in object detection, introducing the concept of combining
region proposals with deep learning-based feature extraction using CNNs. Its accuracy was
unmatched at the time, though its speed was a major limitation. The model inspired faster and
more efficient variants (Fast R-CNN and Faster R-CNN), which have been widely adopted in
real-world applications. R-CNN’s contributions to the field of computer vision, particularly
object detection, continue to influence modern architectures.

SSJCET, Asangaon 6 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

Paper lll

PAPER NAME: CenterNet: Objects as Points.

Author Name: Xiaoming Zhou, Xiaoping Li, Yanjie Wang, Qihua Lin.

Explanation:

Traditional object detection models, like R-CNN, SSD, and YOLO, typically rely on
anchor boxes or predefined bounding boxes to detect objects. Anchor-based methods require
the generation of multiple candidate regions or grid cells to locate objects. This process is
computationally expensive and prone to issues like duplicate detections (non-maximal
suppression) and complex hyperparameter tuning.

To overcome these limitations, CenterNet rethinks object detection by proposing a more

intuitive approach: treating objects as single points. Instead of predicting the bounding boxes
directly, CenterNet detects the center point of each object and then regresses the size and
shape of the bounding box around that point. This strategy reduces the complexity of the
detection pipeline and avoids the problems associated with anchor boxes.

CenterNet introduced a novel and simplified approach to object detection by treating

objects as points and predicting only their centers. This anchor-free, point-based detection
method reduced the complexity of traditional detection pipelines, making the model more
efficient while maintaining competitive accuracy. The simplicity and effectiveness of
CenterNet have made it a significant contribution to the field of real-time object detection,
and it has inspired further research into anchor-free object detection models.

SSJCET, Asangaon 7 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

Comparative Analysis

Table 3.1Comparative Analysis of Existing System

Sr. Author Project Title Publication Technology Purpose

No.
1. Joseph Redmon, You Only CVPR, Convolutional model that can
Santosh Divvala, Look 2016 Neural detect multiple
Ross Girshick, Once Networks objects in an
Ali Farhadi (CNN), Single image in a single
Shot Detection evaluation,
optimizing for
both speed and
accuracy.

2. Ross B. Girshick, R-CNN CVPR, 2014 CNN, Region To improve

Jeff (Region-based Proposal object detection
Donahue, Convolutional Network accuracy by using
Trevor Neural (RPN) regionbased
Networks) convolutional neural
Darrell,
networks.
Jitendra
Malik

3. Xiaoming Zhou, CenterNet: ArXiv Key point CenterNet offers a

Xiaoping Li, Yanjie Objects as preprint, 2019 Estimation, novel
Wang, Qihua Lin Points Heatmapbased approach to
Detection real-time
object
detection by
focusing on the
center points of
objects,
achieving high
efficiency without
sacrificing accuracy.

SSJCET, Asangaon 8 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

4. Existing System
Several existing systems and frameworks are widely used for object detection in
computer vision. Here are some of the popular ones as of my last knowledge update in
January 2022:

1. YOLO (You Only Look Once):

YOLO is a real-time object detection system that can detect objects in images or video
frames in a single pass. It's known for its speed and accuracy. YOLOv4 and YOLOv5 are
some of the latest versions.

2. Faster R-CNN:

Faster Region-based Convolutional Neural Network is a two-stage object detection

framework. It uses region proposals to identify objects and has achieved state-of-the-art
results.

3. SSD (Single Shot MultiBox Detector):

SSD is another single-shot object detection method that is efficient and accurate. It uses
multiple bounding box priors of different scales to detect objects.

4. Retina Net:

Retina Net is designed for accurate object detection. It combines the efficiency of single
shot detectors with the accuracy of two-stage detectors.

5. Mask R-CNN:

Mask R-CNN extends Faster R-CNN by adding a pixel-level mask prediction for each
detected object, allowing instance segmentation in addition to object detection.

6. OpenCV:

OpenCV is an open-source computer vision library that provides pre-trained models for
object detection, including Haar cascades and deep learning-based models.

SSJCET, Asangaon 9 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

5. Problem Statement
The problem statement of object detection is to accurately and efficiently identify
and locate objects within an image or video. The goal is to provide a system that can
automatically detect objects of interest and classify them into different categories, while
also providing bounding box coordinates to precisely locate the objects.

The challenges in object detection include handling variations in object appearance,

dealing with occlusions and cluttered backgrounds, differentiating between similar objects,
and achieving real-time performance. The system should be able to accurately detect and
classify objects in various environments and lighting conditions.

Additionally, the system should be able to detect and locate multiple objects within an
image or video simultaneously, as well as handle objects at different scales, orientations,
and aspect ratios. It should also be robust to changes in viewpoint and object deformations.

The goal is to develop object detection systems that can not only achieve high
accuracy but also operate in real-time or near real-time scenarios, making them applicable
to a wide range of applications such as surveillance, autonomous vehicles, robotics, and
augmented reality.

Overall, the problem statement of object detection is about developing algorithms

and systems that can efficiently and accurately detect, classify, and locate objects in images
or videos, regardless of the complexities and challenges in the input data.

SSJCET, Asangaon 10 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

6. Scope
The scope of object detection encompasses various areas, including computer
vision, machine learning, and artificial intelligence. The field has seen significant
advancements in recent years, with the development of numerous algorithms and
techniques. The scope of object detection includes, but is not limited to, the following
aspects:

1. Algorithm Development: Designing and developing algorithms that can accurately

detect and localize objects in images or videos. This involves exploring and innovating new
techniques to improve the accuracy, speed, and robustness of object detection systems.

2. Data Representation: Representing and encoding both the input data (images or
videos) and the output (object detections). This includes the use of various formats such as
image representations (e.g., RGB, grayscale, etc.), bounding box coordinates, and object
labels or categories.

3. Object Categories: Defining and determining the categories or classes of objects to

be detected. This can range from generic categories like cars, pedestrians, and traffic signs
to more specific and specialized categories in different domains like medical imaging,
industrial inspection, or sports analysis.

4. Training Data and Annotation: Creating and curating high-quality training datasets
with annotated ground truth information, such as bounding box coordinates and object
labels. This involves collecting or generating diverse and representative data to ensure the
effectiveness and generalization of object detection models.

5. Model Selection and Training: Choosing appropriate models and architectures for
object detection, such as Faster R-CNN, SSD, YOLO, or Retina Net. Training these models
involves optimizing the model parameters using large-scale annotated datasets and
appropriate loss functions, such as mean average precision (MAP), to improve detection
accuracy.

6. Evaluation Metrics: Defining and utilizing evaluation metrics to assess the

performance of object detection systems. Common metrics include Intersection over Union
(IOU), precision, recall, MAP, and average precision at different IOU thresholds.

SSJCET, Asangaon 11 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

7. Real-Time and Efficient Implementation: Developing techniques and optimizing

algorithms to achieve real-time object detection, with high processing speed. Efficient
implementation strategies include hardware acceleration, parallel processing, and model
compression techniques.

8. Application Domains: Applying object detection in various domains, such as

autonomous driving, surveillance systems, robotics, healthcare, industrial automation, and
augmented reality. Each domain may present its own challenges and requirements, and the
object detection algorithms should be tailored accordingly.

In summary, the scope of object detection is broad and covers various aspects, techniques,
and applications. It involves algorithm development, data representation, training data,
model selection and training, evaluation, efficient implementation, and application-specific
customization.

SSJCET, Asangaon 12 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

7. Proposed System
 Data Collection and Preprocessing: Gather a diverse dataset of images or videos that
represent the objects you want to detect. Preprocess the data by resizing images,
normalizing pixel values, and augmenting the dataset to increase variability.
 Annotation: Annotate the collected data with bounding box coordinates and object
labels using annotation tools. This step provides ground truth information for training the
object detection model.
 Model Selection: Choose a suitable object detection model based on your
requirements, such as Faster R-CNN, SSD, YOLO, or RetinaNet. Consider factors like
detection accuracy, inference speed, and computational resources available.
 Training: Split the annotated dataset into training and validation sets. Train the object
detection model using the training set, optimizing the model parameters with a suitable
loss function like the focal loss or the smooth L1 loss. Experiment with different
hyperparameters to achieve the best performance.
 Model Evaluation: Evaluate the trained model's performance on the validation set
using evaluation metrics like mAP, precision, and recall. Adjust model hyperparameters or
try different models if necessary to improve performance.
 Model Optimization: Apply optimization techniques to improve the model's efficiency
and speed. For example, you can use model pruning, quantization, or model compression
techniques like knowledge distillation to reduce the model's size without significant loss in
accuracy.

SSJCET, Asangaon 13 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

7.1 System Architecture

Fig. 7.2 System Architecture

 User Interface: This component is responsible for presenting the app's interface to the
user. It includes various screens such as login, search, and results. This component
interacts with the backend via API calls.
 Backend: This component is responsible for processing the user's requests and
generating responses. It includes business logic and communicates with the database
and college API.
 Server: This component runs the backend and serves as the interface between the user
interface and the backend.
 Database: This component stores the app's data, including user data, college data, and
search history.

SSJCET, Asangaon 14 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

7.2 ALGORITHM

import cv2

import torch # for YOLOv8, if using PyTorch model

# Step 1: Load pre-trained YOLOv8 model

model = torch.hub.load('ultralytics/yolov8', 'yolov8')

# Step 2: Set confidence threshold and NMS threshold

conf_thresh = 0.5

nms_thresh = 0.4

# Step 3: Capture video stream (0 = default webcam)

video_capture = cv2.VideoCapture(0)

while True:

# Step 4: Capture each frame from the webcam

ret, frame = video_capture.read()

if not ret:

break

# Step 5: Preprocess the frame (resize, normalize)

preprocessed_frame = cv2.resize(frame, (640, 640))

SSJCET, Asangaon 15 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

# Step 6: Perform object detection on the frame

results = model(preprocessed_frame)

# Step 7: Post-process results (filter by confidence, NMS)

detections = results.xyxy[0].numpy() # Bounding boxes, labels, confidence

for detection in detections:

x1, y1, x2, y2, conf, cls = detection

# Filter detections by confidence

if conf > conf_thresh:

# Draw bounding box

cv2.rectangle(frame, (int(x1), int(y1)), (int(x2), int(y2)), (0, 255, 0), 2)

# Draw label and confidence score

label = f"Class: {int(cls)}, Conf: {conf:.2f}"

cv2.putText(frame, label, (int(x1), int(y1) - 10), cv2.FONT_HERSHEY_SIMPLEX,

0.5, (0, 255, 0), 2)

# Step 8: Display the frame with detected objects

cv2.imshow('Real-Time Object Detection',

frame)

# Step 9: Break loop on 'q' key press

if cv2.waitKey(1) & 0xFF ==

SSJCET, Asangaon 16 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

ord('q'):

SSJCET, Asangaon 17 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

break

# Step 10: Release video capture and close windows

video_capture.release()

cv2.destroyAllWindows()

SSJCET, Asangaon 18 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

 MATHEMATICAL MODEL

1. Object Detection as a Function

Let’s define object detection as a function:
D(x)={(bi,ci,si)}
Where:
 x is an image or frame from a video stream.
 D(x) is the output set of detected objects.
 B represents the bounding box for the i-th object (with coordinates
 cic_ici represents the class label for the i-th object (e.g., car, person, etc.).
 si represents the confidence score for the i-th object (ranging from 0 to 1).

2. Bounding Box Calculation.

The bounding box bi is generally represented by four values corresponding to the

coordinates of the top-left and bottom-right corners of the box. These coordinates are
often predicted as:

bi=(x1,y1,x2,y2)b_i = (x_1, y_1, x_2, y_2)bi=(x1,y1,x2,y2)

These values can be normalized to be relative to the image width and height. For
example, if the image is of size W×H, the normalized bounding box coordinates bi′ ′

3. Confidence Score.

The confidence score sis_isi is the probability of the detected object being the correct
class. It is computed through the classification model’s confidence for the predicted class
ci

si=P (object is class ci∣x)

SSJCET, Asangaon 19 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

4. Loss function

In real-time object detection models, such as YOLO, the total loss is a combination of:

 Classification Loss: Cross-entropy loss for the predicted class labels.

 Localization Loss: Mean Squared Error (MSE) for the predicted bounding box
coordinates.
 Confidence Loss: Measures how confident the model is about the presence of an object.

SSJCET, Asangaon 11 Department of Computer Engg.

0
Mini Project - 1 A Real Time Object Detection

Fowchart

fig.8 1Flowchart

SSJCET, Asangaon 20 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

8. Planning

fig.9 1Gantt chart for planning

SSJCET, Asangaon 21 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

9. Designing details
9.1 Coding
import cv2

import numpy as np

# Load YOLO

model def

load_yolo():

net = cv2.dnn.readNet("yolov3.weights", "yolov3.cfg")

with open("coco.names", "r") as f:

classes = [line.strip() for line in f. readlines()]

layer_names = net.getLayerNames()

output_layers = [layer_names[i - 1] for i in net.getUnconnectedOutLayers()]

return net, classes, output_layers

# Perform object detection on a frame

def detect_objects(imgm, net, output_layers):

height, width = img.shape[:2]

blob = cv2.dnn.blobFromImage(img, 0.00392, (416, 416), (0, 0, 0), True, crop=False)

net.setInput(blob)

detections = net.forward(output_layers)

boxes = []

confidences = []

class_ids = []

for output in detections:

SSJCET, Asangaon 22 Department of Computer Engg.
Mini Project - 1 A Real Time Object Detection

for detection in output:

scores = detection[5:]

class_id = np.argmax(scores)

confidence = scores[class_id]

if confidence > 0.5:

center_x = int(detection[0] * width)

center_y = int(detection[1] * height)

w = int(detection[2] * width)

h = int(detection[3] * height)

x = int(center_x - w / 2)

y = int(center_y - h / 2)

boxes.append([x, y, w, h])

confidences.append(float(confidence))

class_ids.append(class_id)

return boxes, confidences, class_ids

# Draw bounding boxes on the frame

def draw_labels(img, boxes, confidences, class_ids, classes):

indexes = cv2.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)

SSJCET, Asangaon 23 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

font = cv2.FONT_HERSHEY_PLAIN

colors = np.random.uniform(0, 255, size=(len(classes), 3))

if len(indexes) > 0:

for i in indexes.flatten():

x, y, w, h = boxes[i]

label = str(classes[class_ids[i]])

confidence = str(round(confidences[i], 2))

color = colors[class_ids[i]]

cv2.rectangle(img, (x, y), (x + w, y + h), color, 2)

cv2.putText(img, label + " " + confidence, (x, y - 5), font, 1, color, 2)

# Main function for real-time object detection

def start_video_detection():

net, classes, output_layers = load_yolo()

cap = cv2.VideoCapture(0) # Use 0 for webcam, or replace with video file path

while True:

ret, frame = cap.read()

if not ret:

break

# Perform detection

boxes, confidences, class_ids = detect_objects(frame, net, output_layers

# Draw the bounding boxes

SSJCET, Asangaon 24 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

draw_labels(frame, boxes, confidences, class_ids, classes)

# Display the frame

cv2.imshow("Real-Time Object Detection", frame)

if cv2.waitKey(1) & 0xFF == ord('q'): # Press 'q' to quit

break

cap.release()

cv2.destroyAllWindows()

# Run the detection

if name == " main ":

start_video_detection()

SSJCET, Asangaon 25 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

Screenshots

SSJCET, Asangaon 26 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

10. Conclusion
Real-time object detection is a transformative technology that has applications in various
fields such as surveillance, autonomous vehicles, robotics, and augmented reality. By
leveraging advanced deep learning models like YOLO, real-time object detection allows
systems to identify and track objects in live video streams with high accuracy and speed.

The integration of pre-trained models with efficient algorithms enables detection systems to
process images and videos in real time, making them suitable for practical deployment in
dynamic environments. Despite challenges like handling occlusion, varying lighting
conditions, and real-time performance constraints, continuous improvements in neural
network architectures and hardware acceleration (such as GPUs) are pushing the boundaries
of what’s possible in this field.

In summary, real-time object detection offers a powerful combination of accuracy and speed,
which is essential for developing intelligent systems capable of interacting with and
interpreting the visual world as it happens.

SSJCET, Asangaon 27 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

11. Requirements
11.1 Hardware Requirements
System : Intel(R) Core (TM) i5-9300H CPU @ 2.40GHz

Hard disk : 40 GB.

Floppy Drive : 1.44 Mb.

Monitor : 14’ Color Monitor.

Mouse : Optical Mouse.

Ram : 2 GB.

Keyboard : 101 Keyboard Keys.

11.2 Software Requirements

Operating system : Window 10

Coding Language : Python

Software’s Used : JDK 1.7, eclipse

Database : MySQL

SSJCET, Asangaon 28 Department of Computer Engg.

Mini Project - 1 A Real Time Object Detection

12. References
[1] Analogy. Wikipedia, Mar 2018.

[2] M. Everingham, L. Van Gool, C. K. Williams, J. Winn, and A. Zisserman. The pascal
visual object classes (voc) challenge. International journal of computer vision, 88(2):303–
338, 2010.

[3] A. Shrivastava, R. Sukthankar, J. Malik, and A. Gupta. Beyond skip connections: Top
down modulation for object detection. arXiv preprint arXiv:1612.06851, 2016.

[4] I. Krasin, T. Duerig, N. Alldrin, V. Ferrari, S. Abu-El-Haija, A. Kuznetsova, H. Rom, J.

Uijlings, S. Popov, A. Veit, S. Belongie, V. Gomes, A. Gupta, C. Sun, G. Chechik, D. Cai, Z.
Feng, D. Narayanan, and K. Murphy. Openimages: A public dataset for large-scale multilabel
and multi-class image classification. Dataset available from https://github.com/openimages,
2017.

[5] T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollar, and C. L.
Zitnick. Microsoft coco: Com- ´ mon objects in context. In European conference on computer
vision, pages 740–755. Springer, 2014.

[6] T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollar, and C. L.
Zitnick. Microsoft coco: Com- ´ mon objects in context. In European conference on computer
vision, pages 740–755. Springer, 2014.

[7] J. Redmon and A. Farhadi. Yolo9000: Better, faster, stronger. In Computer Vision and
Pattern Recognition (CVPR), 2017 IEEE Conference on, pages 6517–6525. IEEE, 2017. 1, 2,
3 [16] J. Redmon and A. Farhadi. Yolov3: An incremental improvement. arXiv, 2018.

[8] O. Russakovsky, L.-J. Li, and L. Fei-Fei. Best of both worlds: human-machine
collaboration for object annotation. In Proceedings of the IEEE Conference on Computer
Vision and Pattern Recognition, pages 2121–2131, 2015.

SSJCET, Asangaon 29 Department of Computer Engg.