Scribd
Upload
3 views4 pages

Proj Milestones, Description

The document outlines a project for a Traffic Sign Recognition and Detection System aimed at identifying and classifying Pakistan-specific traffic signs from live video feeds. The project includes milestones for dataset collection, model training, implementation, and documentation, with a total budget of Rs 13,000 and a timeline of 11 days. Performance metrics indicate an inference speed of 10-15 FPS, with classification accuracies of 86% for mandatory signs, 82% for warnings, and 78% for informatory signs.

Uploaded by

moena0661
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
3 views4 pages

Proj Milestones, Description

The document outlines a project for a Traffic Sign Recognition and Detection System aimed at identifying and classifying Pakistan-specific traffic signs from live video feeds. The project includes milestones for dataset collection, model training, implementation, and documentation, with a total budget of Rs 13,000 and a timeline of 11 days. Performance metrics indicate an inference speed of 10-15 FPS, with classification accuracies of 86% for mandatory signs, 82% for warnings, and 78% for informatory signs.

Uploaded by

moena0661
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

PROJECT MILESTONES

“ TRAFFIC SIGN RECOGNITION AND DETECTION SYSTEM ”

PROJECT DESCRIPTION

A cost-effective, short-term solution to detect and classify Pakistan-specific traffic signs


from live video feeds. The system focuses on:
 Detection: Identify signs in near-real-time (10–15 FPS) from video inputs (no
strict 30 FPS requirement).
 Classification: Categorize signs into Mandatory, Warning,
and Informatory classes.
 Output: Display labels with descriptions (e.g., "Speed Limit 40 km/h") for driver
assistance.

Scope Adjustments:
 Simplified model architecture (YOLOv8 or MobileNet for speed-accuracy trade-
off).
 Limited dataset size (1,000–1,500 images) with augmentation.

MILESTONES, TIMELINES AND PAYMENT SCHEDULE


Total Budget: Rs 13,000 | Timeline: 11 Days

Ser Milestone Objectives Deliverables Timeline Payment

Collect 1,000
Pakistan-specific Raw dataset (Google
Dataset traffic sign images. Drive link).
Rs 3,000
1. Formulation 3 Days
(Advance)
(Advance) Basic annotations Annotation files (CSV
(bounding boxes + or COCO format).
labels).
Train lightweight
detection/classifica
Trained model weights.
tion model.
2. Model Training 3 Days Rs 3,000
Accuracy/validation
Achieve ≥85%
report.
validation
accuracy.
Demo video with
Integrate model
detections (10–15
with live video
Implementation & FPS).
3. input (OpenCV). 3 Days Rs 4,000
Testing
Test on 5+ real-
Test results
world videos.
(precision/recall).
 Document:
Methodology (model
architecture, tools).

Experiments
(hyperparameters,
Prepare detailed
augmentations).
Final report with

4. Documentation methodology, 2 Days Rs 3,000
Results (accuracy, F1,
(Final Payment) results, and
confusion matrix).
graphs.
Testing validation
(sample outputs).

Conclusion.
Graphs (accuracy
curves, class-wise F1).

RESULTS

1. Performance Metrics

 Inference Speed: 10–15 FPS (tested on mid-range laptop CPU).


 Latency: ~60–100 ms per frame.

2. Classification Results (Example)

Category Accuracy F1 Score


Mandatory 86% 0.84
Warning 82% 0.79
Informatory 78% 0.75

3. Confusion Matrix (Example)

Actual \ Predicted Speed Limit 40 Road Bump School Zone


Speed Limit 40 85% 10% 5%
Road Bump 8% 80% 12%
School Zone 3% 15% 82%
4. Label Examples
Class Description
SL-40 Mandatory: Speed Limit 40 km/h
RB-01 Warning: Road Bump Ahead
SZ-01 Informatory: School Zone

5. Documentation Structure

 Introduction: Project scope and objectives.


 Methodology:
o Tools: Python, OpenCV, Ultralytics (YOLOv8).
o Model: YOLOv8 for detection and classification.
 Experiments:
o Data augmentation (rotation, noise).
o Hyperparameters (epochs=50, batch_size=8).
 Results:
o Graph 1: Training vs. Validation Accuracy.
o Graph 2: Class-wise F1 Scores.
o Table: Comparison with baseline models.
 Testing: Sample video frame outputs.
 Conclusion: Limitations and future work.
 Future Scope:
o Real-Time Optimization: Deploy on edge devices (NVIDIA Jetson) for 30
FPS performance.
o Dataset Expansion: Include more Pakistan-specific signs and diverse
conditions (night, rain).
o Advanced Models: Experiment with YOLOv9 or transformer-based models
(DETR).
o Integration: Embed system into ADAS (Advanced Driver Assistance
Systems) or traffic monitoring systems.
o User Interface: Develop a mobile app for real-time sign detection and alerts.
6. Notes
 Payment release is milestone-based.
 Dataset quality depends on accessibility of Pakistan-specific sign images.
 Documentation will include graphs and tables for clarity.

You might also like