ev_objdet_pc

An object detection task with only persons and cars in bounding boxes (COCO format)

---

Training

git clone https://github.com/partham16/ev_objdet_pc.git
cd ev_objdet_pc
python3.8 -m venv pyenv38
source pyenv38/bin/activate
pip install -r requirements.txt
python ev_train.py

or use make full_install for full developmental set up.

Note:

in src/config.py the image_bs is set as 24 - that might cause the memory to be exhausted - reduce the batch size in that case.

Check out Model Prediction

Check out Model demo - WIP

A docker container deployed on Azure Container Instances

• Caveat:

  .png images generally aren't supported - use .jpg

  as stated, WIP, certain edge cases around display needs to be looked into

Read more on design choices

README_ML.md