YOLOv2 in PyTorch
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 47 commits ahead, 36 commits behind longcw:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
cfgs
datasets
demo
flow
layers
misc
train
train_data
utils
.gitignore
README.md
darknet.py
darknet_training_v3.py
darknet_v3.py
make.sh
yolo_detect.py

README.md

YOLOv2 in PyTorch

This is a PyTorch implementation of YOLOv2. This project is forked from (https://github.com/longcw/yolo2-pytorch), but not compatible with origin version.

Currently, I train this model for KITTI Dataset to demo. It predicts car, pedestrian and cyclist. If you want a general detecotr, please refer to this.

You can also use original YOLOv2 COCO model on KITTI, Here is a demo video

For details about YOLO and YOLOv2 please refer to their project page and the paper: YOLO9000: Better, Faster, Stronger by Joseph Redmon and Ali Farhadi.

System Environment

  • Ubuntu 16.04
  • CUDA 8.0 / cuDNN 5.1
  • Python 3.5
  • Numpy 1.12
  • PyTorch 0.1.12
  • OpenCV 3.2

With a 1080Ti GPU, I get ~30 fps using this KITTI model (input size = 1216 x 352)

Installation and demo

  1. Clone this repository

    git clone git@github.com:cory8249/yolo2-pytorch.git
  2. Build the reorg layer (tf.extract_image_patches)

    cd yolo2-pytorch
    ./make.sh
  3. Download the trained model kitti_baseline_v3_100.h5 and set the model path in yolo_detect.py

  4. Run demo python3 yolo_detect.py.

Install any missing packages manually via pip