BOSCH_A-G_INTERIIT

Repository containing Our Solution for BOSCH Age and Gender Detection Problem Statement in InterIIT TechMeet 10.0

A snapshot for our approach is shown below : -

INSTALLATION

Step1 - Cloning this Repository

• In terminal/git bash run

git clone https://github.com/Amshra267/BOSCH_A-G_INTERIIT.git
## move to that directory
cd BOSCH_A-G_INTERIIT

Step 2 - We had provided two configurations for installation

• For GPU(Recommended) - (Requirements - CUDA version = 11.1)

  • In the above opened terminal run

  chmod +x install_gpu.bash
  ./install_gpu.bash

• For CPU

  • In the above opened terminal run

  chmod +x install_cpu.bash
  ./install_cpu.bash

Now you are ready to run our solution

TESTING

In terminal run

python main.py --video <path_to_your_video_file>

Above file have some arguments:-

--video = <path of video file> 
             or <camera id_no in case of real time image feed
            example = 0 (for webcam ), 1 (for external camera) >

--show  = 1 (for showing our output), default = 0  

--output = <path_of_output_saved_video_result>

OUTPUT FORMAT

We are creating three folders inside the outputs folder corresponding to the passed video's as an input result, persons in their cropped form and labels file having their demographic details.

The output inside the labels.txt is in the format as follows; Timestamp, Person Id, Gender , Age Range , Age

we added a test file for you in outputs/video_results folder

DATASET USED

• We trained Age and Gender from a Gait Rcognition Dataset TUM-GAID DATASET which we got from a CS Department Professor and took authorization from him.
• Trained UNet on Open Source data - OCHuman Dataset
• Other models and datasets are available open source like coco for object detection. Mars dataset.

References

• Yolov4 (For Object Detection) - https://arxiv.org/abs/2004.10934
• Deep-Sort (For Object Tracking) - https://arxiv.org/abs/1703.07402
• ESRGAN (For super resolution) - https://arxiv.org/abs/1809.00219
• UNet (For Segmentation Mask) - https://arxiv.org/abs/1505.04597
• Gait Recognition (For Age and Gender analysis) - https://link.springer.com/10.1007/978-1-4419-5906-5_741