A Computer Vision Framework to run CV models and tasks.
- Human Activity Recognition
- Object Detection
- Pose Estimation
This project was mainly based on:
- For project architecture and action recognition: Kenshohara et al.
- For object detection (Yolov4 - Adapted): Tianxiaomo (Original): AlexeyAB
- For pose estimation (Open Pose - Adapted): Hzzone (Original): Hidalgo, Cao and Simon - CMU
- Clone repository
- Create your own virtual environment (Optional but highly recommended)
python -m venv myenv
- Activate your new environment (Optional but highly recommended)
myenv/scripts/activate
- Install python required libraries
python pip install -r requirements.txt
- Install pytorch (pip installation is failing yet)
python pip install torch===1.6.0 torchvision===0.7.0 -f https://download.pytorch.org/whl/torch_stable.html
- Download model's weights for demo
- Yolov4 - Object Detection
- Save in /models/obj_det
- Body - Pose Estimation
- Save in /models/pose
- Hand - Pose Estimation
- Save in /models/pose
- Yolov4 - Object Detection
There are two ways to set your options to run the code.
- Passing it in command line.
python main.py --opts_dst_path opts/opts_output.json
- Setting it in the opts.json file and add the path to this file in the arguments.
python main.py --opts_src_path opts/opts_input.json
Use the code below to test your webcam read.
python main.py --webcam
Use the code below to demo the OD algorithm with your webcam input.
python main.py --webcam --od
Use the code below to demo the PE body algorithm with your webcam input.
python main.py --webcam --pe
Use the code below to demo the PE body and hands algorithm with your webcam input.
python main.py --webcam --pe --pe_hand True
Use the code below to demo the Action Recognition algorithm with your webcam input.
python main.py --webcam --ar
- Guilherme Augusto Silva Surek
- Mateus Isaac Di Domenico
- Matheus Henrique Reis Marchioro