packyan / PyTorch-YOLOv3-kitti

use yolov3 pytorch to train kitti

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool


Minimal implementation of YOLOv3 in PyTorch. And Training from Kitti dataset

Table of Contents

HI THERE~ THIS repo is forked from eriklindernoren


$ git clone
$ cd PyTorch-YOLOv3-kitti/
$ sudo pip3 install -r requirements.txt
Download pretrained weights

if you wan use pretrained darknet-53 on IMAGENET weights, please download darknet53.conv.74,and put it into checkpoints/

if you just want a pretrained weights on kitti dataset for test or detect, please download pretrained weights file, and put it into weights folder, the path: weights/yolov3-kitti.weights

Download Kitti

The KITTI Vision Benchmark Suite

and you should transfrom kitti lable to coco label, by using label_transform


Uses pretrained weights to make predictions on images. weights/yolov3-kitti.weights was trained by kitti data set. python3 --image_folder /data/samples

Small objects detection

Detect to detect objects, and please put samples into data/samples defult weights files is weights/kitti.weights


run to detect objects from a webcam or a video file. On I7 7700 8G GTX1070 FPS is 22 cause some problems, test resized images is really faster than resizeing images and then inference.




Please run python3 -m visdom.server first to vislizer your training loss.

Data augmentation as well as additional training tricks remains to be implemented. PRs are welcomed! [-h] [--epochs EPOCHS] [--image_folder IMAGE_FOLDER]
                [--batch_size BATCH_SIZE]
                [--model_config_path MODEL_CONFIG_PATH]
                [--data_config_path DATA_CONFIG_PATH]
                [--weights_path WEIGHTS_PATH] [--class_path CLASS_PATH]
                [--conf_thres CONF_THRES] [--nms_thres NMS_THRES]
                [--n_cpu N_CPU] [--img_size IMG_SIZE]
                [--checkpoint_interval CHECKPOINT_INTERVAL]
                [--checkpoint_dir CHECKPOINT_DIR]


YOLOv3: An Incremental Improvement

Joseph Redmon, Ali Farhadi

We present some updates to YOLO! We made a bunch of little design changes to make it better. We also trained this new network that’s pretty swell. It’s a little bigger than last time but more accurate. It’s still fast though, don’t worry. At 320 × 320 YOLOv3 runs in 22 ms at 28.2 mAP, as accurate as SSD but three times faster. When we look at the old .5 IOU mAP detection metric YOLOv3 is quite good. It achieves 57.9 AP50 in 51 ms on a Titan X, compared to 57.5 AP50 in 198 ms by RetinaNet, similar performance but 3.8× faster. As always, all the code is online at

[Paper] [Original Implementation]


  title={YOLOv3: An Incremental Improvement},
  author={Redmon, Joseph and Farhadi, Ali},
  journal = {arXiv},


use yolov3 pytorch to train kitti


Language:Python 95.5%Language:Jupyter Notebook 4.4%Language:Shell 0.1%