Scribd Logo
Upload

Welcome to Scribd!

  • 13 views

    AI Engineer Hiring Assignment

    Uploaded by

    VISHNU VICKY
    The document outlines 4 tasks for an AI engineering assignment involving computer vision and NLP, including developing code to detect object outlines, draw dots on polka dots in a video, run EasyOCR on license images, and do pose estimation on a video using MediaPipe. Candidates are instructed to complete the tasks, record demos, and submit their code and output files to a shared Google Drive folder for evaluation.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    AI Engineer Hiring Assignment

    Uploaded by

    VISHNU VICKY
    13 views5 pages
    The document outlines 4 tasks for an AI engineering assignment involving computer vision and NLP, including developing code to detect object outlines, draw dots on polka dots in a video, run EasyOCR on license images, and do pose estimation on a video using MediaPipe. Candidates are instructed to complete the tasks, record demos, and submit their code and output files to a shared Google Drive folder for evaluation.

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document outlines 4 tasks for an AI engineering assignment involving computer vision and NLP, including developing code to detect object outlines, draw dots on polka dots in a video, run EasyOCR on license images, and do pose estimation on a video using MediaPipe. Candidates are instructed to complete the tasks, record demos, and submit their code and output files to a shared Google Drive folder for evaluation.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    13 views5 pages

    AI Engineer Hiring Assignment

    Uploaded by

    VISHNU VICKY
    The document outlines 4 tasks for an AI engineering assignment involving computer vision and NLP, including developing code to detect object outlines, draw dots on polka dots in a video, run EasyOCR on license images, and do pose estimation on a video using MediaPipe. Candidates are instructed to complete the tasks, record demos, and submit their code and output files to a shared Google Drive folder for evaluation.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 5

    AI Engineer - Computer Vision & NLP - Full Time Hiring Assignment

    Read “Assignment Submission” at the end of this document before


    proceeding with the assignment.

    Note - The Assignment will take 2 to 3 hrs to complete. Many


    candidates won’t be able to complete the task. Please take up this
    assignment as a fun challenge!

    Watch the video “Assignment Explanation.mp4” in the drive. It


    contains the explanation of the assignment.

    1. Task 1: Automatic Object Outline/Border Detection


    (Development) (15 Points)
    Develop a basic OpenCV Python script that draws an object’s outline
    using rembg library & OpenCV Image processing.
    a. In the Python script add a variable to specify the input
    image path.
    b. The script should read the image & display it.
    c. Ask the user to draw a rectangle over an object inside the
    image. (cv2.selectROI)
    d. Crop the selected Region Of Interest (ROI) and pass it to
    rembg library. (https://github.com/danielgatis/rembg) [pip
    install rembg]
    e. The library will remove the background from the ROI and
    return the image back.
    f. Draw an outline over the object since it doesn’t have any
    background now.
    g. Overlay the object outline in the original input image and
    show it as output.
    h. Allow the user to do this again and again until “q” is
    pressed.
    i. To clear the outline, allow the user to press “c”

    Submission -
    j. Screen record using Loom/OBS/Any Other Software and
    run your script on “1.jpg” & “2.jpg” in the “TEST IMAGES”
    folder.
    k. Draw a box over the women in red pants for “1.jpg”
    l. Draw a box over the car for “2.jpg”

    2. Task 2: Draw Red Dot On Green Polka Dots (Development) (10


    Points)
    a. The input video for this task is “task_2_video.mp4”
    b. In this task, detect the green polka dots in the video. (HSV
    filtering)
    c. You can hardcode the HSV range values for the green color.
    d. Draw a red dot at the center of each green polka dot.
    e. Do this for all the frames in the video and generate an
    output video.
    f. You can consider both the green dots (Dark & Light) or any
    one of them.
    g. You can keep the size of the red dot the same for all the
    green dots.
    h. Output example of a single frame -
    i. Write an algorithm to output such red dots drawn at the
    center of the green dots.
    j. If 2 green dots overlap, you can draw a single red dot or 2
    whichever you feel is easier for you.

    3. Task 3: EasyOCR (Development) (7 Points)


    a. For this task, use the EasyOCR library in Python.
    b. Write a program in Python OpenCV to read all the image
    files in a folder.
    c. For each file, run the EasyOCR library and store all the
    output images in a folder
    d. Input images will look like “task_3_input.jpg”
    e. The output image should look like
    “task_3_partial_output.png” but in place of red boxes with
    numbers, it should be red boxes with green-colored OCR
    detected text beside it. Example is given for “DOB
    03/05/1960” in the “task_3_partial_output.png”
    f. Develop this algorithm and run it on all the images inside
    “License Images.zip”.
    g. Store all the output images in a folder named “License
    Outputs” and upload the folder to your submission drive
    folder. Do not zip the “License Outputs” folder. The images
    should be viewable directly on Google Drive.
    h. Upload your .py file to your Google Drive folder.

    4. Task 4: MediaPipe Pose Estimation (Inference) (5 Points)


    a. The input video for this task is “task_4_video.mp4”
    b. Draw the person’s skeleton in the video using
    MediaPipe Library
    c. This task is just to check if you’re able to use MediaPipe
    pose estimation for inference.
    d. This is a pose estimation task and not face key point
    detection.

    Assignment Submission -

    1. Your shared folder in Google Drive will contain the following -


    a. Task 1 -
    i. “task_1_code.py” - Python code for border detection.
    ii. “task_1_demo.mp4” - Screen recording of the script
    running on “TEST IMAGES”
    b. Task 2 -
    i. “task_2_output.mp4” - Output video generated by
    running the polka dots detection script
    ii. “task_2_code.py” - Python code for polka dots detection
    c. Task 3 -
    i. “License Outputs” - This folder will contain the output
    images for the EasyOCR task.
    ii. “code_task_3.py” - Python code for EasyOCR Task
    d. Task 4 -
    i. “output_task_4.mp4” - Output video of MediaPipe task
    “code_task_4.py” - Python code for MediaPipe Task
    ii.
    2. Do NOT zip the files and upload them. Assignments that are
    zipped won’t be considered.
    3. The Google Drive folder link should be set to "Anyone with the
    link can view". Without this, your submission won't be accepted.

    All the best!

    You might also like