tqjxlm/deepcrowd-pytorch

DeepCrowd

Experiment implementation of DeepCrowd.

Usage

New train

Modify configs/new-train.json, then run

python main.py new-train [task number]

Load and resume a previous train

Modify configs/resume.json, then run

python main.py resume [task number]

Inference and render

Modify configs/resume.json, then run

python main.py inference [task number]

Change hyper-paramters

Most key hyper-paramters are specified in configs/*.json

Other parameters are specified in config.py

Change task parameter

Task parameters are specified in challenges/*.json

Other stage parameters are specified in env/stage.py

Dependencies

• pytorch 1.0
• pyqtgraph
• tensorboard