packyan / pytorch-yolov3-kitti Goto Github PK

View Code? Open in Web Editor NEW

121.0 3.0 63.0 39.89 MB

use yolov3 pytorch to train kitti

Python 95.54% Jupyter Notebook 4.35% Shell 0.11%

pytorch-yolov3-kitti's People

Contributors

Stargazers

Watchers

pytorch-yolov3-kitti's Issues

train/val split used to obtain yolov3-kitti.weights

Thanks for sharing your work.
Could you please clarify the train/val split that was used to train the model corresponding to the yolov3-kitti.weights?
Is it the standard 3717 / 3769 split or was all of the 7481 images used for training?

Why are the AP scores all 0 when running test.py?

I'm using the pretrained kitti weights.
I'm testing on the KITTI 2d dataset downloaded from the official site.
I converted the kitti labels to coco format and put them under PyTorch-YOLOv3-kitti\data\kitti\labels\test
Test images are under PyTorch-YOLOv3-kitti\data\kitti\image\test

pretrained kitti weight epoch?

Hi Packyan,
Thanks for the code. When was the weights are saved? Is it at default 100 epoch or some other?

Format for targets

Hi!

I just wanted to ask about what is the final format of the targets when you pass them to the network for training?

From your code I understand that you read the GT from a txt file, but on datasets.py in line 96 you use np.readtxt but the labels begin with a string (see this example).

Could you please do me the favor of specifying what is the composition of the targets? I know that per image you limit the amount of boxes to 50, so the targets will have shape [batch_size, 50, 5]

Are they in the form [class, center_x_ratio, center_y_ratio, box_width_ratio, box_height_ratio]?

Where ratio means the value in pixels divided by the width of the image (for center_x and box_width) or by the height of the image (for center_y and box_height).

Thank you!

Training on Custom Dataset

Hi,

First of all thank you for the code and detailed steps involved.
I was trying to run the same code with another dataset (Virtual KITTI). I am able to start the training process but I am not sure how to decide the number of epochs. Is there any way to implement early stopping in the code or what should be the process. Can you please guide

plt.get_cmap uses vega20b which doesn't exist

On executing detect.py if Vega20b is used as color map,an errror is thrown of the following form

Vega20b' is not a valid value for name; supported values are 'Accent', 'Accent_r', 'Blues', 'Blues_r', 'BrBG', 'BrBG_r', 'BuGn', 'BuGn_r', 'BuPu', 'BuPu_r', 'CMRmap', 'CMRmap_r', 'Dark2', 'Dark2_r', 'GnBu', 'GnBu_r', 'Greens', 'Greens_r', 'Greys', 'Greys_r', 'OrRd', 'OrRd_r', 'Oranges', 'Oranges_r', 'PRGn', 'PRGn_r', 'Paired', 'Paired_r', 'Pastel1', 'Pastel1_r', 'Pastel2', 'Pastel2_r', 'PiYG', 'PiYG_r', 'PuBu', 'PuBuGn', 'PuBuGn_r', 'PuBu_r', 'PuOr', 'PuOr_r', 'PuRd', 'PuRd_r', 'Purples', 'Purples_r', 'RdBu', 'RdBu_r', 'RdGy', 'RdGy_r', 'RdPu', 'RdPu_r', 'RdYlBu', 'RdYlBu_r', 'RdYlGn', 'RdYlGn_r', 'Reds', 'Reds_r', 'Set1', 'Set1_r', 'Set2', 'Set2_r', 'Set3', 'Set3_r', 'Spectral', 'Spectral_r', 'Wistia', 'Wistia_r', 'YlGn', 'YlGnBu', 'YlGnBu_r', 'YlGn_r', 'YlOrBr', 'YlOrBr_r', 'YlOrRd', 'YlOrRd_r', 'afmhot', 'afmhot_r', 'autumn', 'autumn_r', 'binary', 'binary_r', 'bone', 'bone_r', 'brg', 'brg_r', 'bwr', 'bwr_r', 'cividis', 'cividis_r', 'cool', 'cool_r', 'coolwarm', 'coolwarm_r', 'copper', 'copper_r', 'cubehelix', 'cubehelix_r', 'flag', 'flag_r', 'gist_earth', 'gist_earth_r', 'gist_gray', 'gist_gray_r', 'gist_heat', 'gist_heat_r', 'gist_ncar', 'gist_ncar_r', 'gist_rainbow', 'gist_rainbow_r', 'gist_stern', 'gist_stern_r', 'gist_yarg', 'gist_yarg_r', 'gnuplot', 'gnuplot2', 'gnuplot2_r', 'gnuplot_r', 'gray', 'gray_r', 'hot', 'hot_r', 'hsv', 'hsv_r', 'inferno', 'inferno_r', 'jet', 'jet_r', 'magma', 'magma_r', 'nipy_spectral', 'nipy_spectral_r', 'ocean', 'ocean_r', 'pink', 'pink_r', 'plasma', 'plasma_r', 'prism', 'prism_r', 'rainbow', 'rainbow_r', 'seismic', 'seismic_r', 'spring', 'spring_r', 'summer', 'summer_r', 'tab10', 'tab10_r', 'tab20', 'tab20_r', 'tab20b', 'tab20b_r', 'tab20c', 'tab20c_r', 'terrain', 'terrain_r', 'twilight', 'twilight_r', 'twilight_shifted', 'twilight_shifted_r', 'viridis', 'viridis_r', 'winter', 'winter_r'

The error is resolved if I replace Vega20b with tab20b.Maybe it's because later versions of matplotlib don't support vega20b.

KITTI training performance

Hi packyan,
The KITTI val performance of using the pre-trained model is very nice, which is:
Average Precisions:

Class '0' - AP: 0.8874476379466757
Class '1' - AP: 0.8776504929104587
Class '2' - AP: 0.9094174650326788
Class '3' - AP: 0.7935572902584845
Class '4' - AP: 0.6938401233821029
Class '5' - AP: 0.7484522134591276
Class '6' - AP: 0.9765143493123802
Class '7' - AP: 0.6567834290009646
mAP: 0.8179578751628592

However, when doing the training with the Imagenet pretrained-model, I am only able to get the following results, which is:
Average Precisions:

Class '0' - AP: 0.5895481840925929
Class '1' - AP: 0.056750284687503164
Class '2' - AP: 0.05041402945683988
Class '3' - AP: 0.25446338964588194
Class '4' - AP: 0.0
Class '5' - AP: 0.042098299062717656
Class '6' - AP: 0.0
Class '7' - AP: 0.0
mAP: 0.12415927336819195
�
I'm wondering what thing do I miss or configure? I use the same label-transforming pre-processing code as in the instruction and training without changing the python arguments. Is anything wrong I did or any arguments need to be changed?

Thanks a lot!

Kitti.weights

Hi
What "kitti.weights" ? And how to download it?
Please help me

Train.py and kitti labels

Hi,

So I am using my own data from http://www.cvlibs.net/datasets/kitti/raw_data.php?type=residential
I am using the second to last data set 2011_10_03_drive_0027 (17.6 GB), I am able to run all necessary code until I get to the train.py file I keep getting the same error of

train.py", line 247
for i in tqdm.tqdm(range(len(all_annotations)), desc=f"Computing AP for class '{label}'"):
^
SyntaxError: invalid syntax

and if not that other ones that I will post once I get them again, also my visdom screen stays blue the whole time, and I'm not sure how to fix that issue. I've looked on StackOverflow and GitHub nothing resolves the issue.
I also do not know how to obtain labels for the dataset I am using can anyone help, please.

RuntimeError: CUDA out of memory

I'm getting RuntimeError for GPU even after I have set batch size to 1 instead of mentioned in repo 16.
Could you please guide me into this for any possible correction?
GPU is having capacity of 8GB

Thank you!

运行video.py， fps很低

我测试个视频，fps只有2，比单独测试图片慢很多，不知你有没测试过

pretrained kitti weight

Download pretrained weights please download weights/yolov3-kitti.weights

Thank you for your code repository. I saw the trained weight of the Kitti in your readme to be download. Where can I download it?

运行python3 detect.py --image_folder /data/samples时报错，求解答

Config:
Namespace(batch_size=1, class_path='data/kitti.names', conf_thres=0.8, config_path='config/yolov3-kitti.cfg', image_folder='/data/samples', img_size=416, n_cpu=8, nms_thres=0.4, use_cuda=True, weights_path='weights/kitti.weights')
/usr/local/lib/python3.5/dist-packages/torch/nn/_reduction.py:43: UserWarning: size_average and reduce args will be deprecated, please use reduction='mean' instead.
warnings.warn(warning.format(ret))
model path: weights/kitti.weights
data size : 0

Performing object detection:
Traceback (most recent call last):
File "detect.py", line 90, in
cmap = plt.get_cmap('Vega20b')
File "/usr/local/lib/python3.5/dist-packages/matplotlib/cm.py", line 182, in get_cmap
% (name, ', '.join(sorted(cmap_d))))
ValueError: Colormap Vega20b is not recognized. Possible values are: Accent, Accent_r, Blues, Blues_r, BrBG, BrBG_r, BuGn, BuGn_r, BuPu, BuPu_r, CMRmap, CMRmap_r, Dark2, Dark2_r, GnBu, GnBu_r, Greens, Greens_r, Greys, Greys_r, OrRd, OrRd_r, Oranges, Oranges_r, PRGn, PRGn_r, Paired, Paired_r, Pastel1, Pastel1_r, Pastel2, Pastel2_r, PiYG, PiYG_r, PuBu, PuBuGn, PuBuGn_r, PuBu_r, PuOr, PuOr_r, PuRd, PuRd_r, Purples, Purples_r, RdBu, RdBu_r, RdGy, RdGy_r, RdPu, RdPu_r, RdYlBu, RdYlBu_r, RdYlGn, RdYlGn_r, Reds, Reds_r, Set1, Set1_r, Set2, Set2_r, Set3, Set3_r, Spectral, Spectral_r, Wistia, Wistia_r, YlGn, YlGnBu, YlGnBu_r, YlGn_r, YlOrBr, YlOrBr_r, YlOrRd, YlOrRd_r, afmhot, afmhot_r, autumn, autumn_r, binary, binary_r, bone, bone_r, brg, brg_r, bwr, bwr_r, cividis, cividis_r, cool, cool_r, coolwarm, coolwarm_r, copper, copper_r, cubehelix, cubehelix_r, flag, flag_r, gist_earth, gist_earth_r, gist_gray, gist_gray_r, gist_heat, gist_heat_r, gist_ncar, gist_ncar_r, gist_rainbow, gist_rainbow_r, gist_stern, gist_stern_r, gist_yarg, gist_yarg_r, gnuplot, gnuplot2, gnuplot2_r, gnuplot_r, gray, gray_r, hot, hot_r, hsv, hsv_r, inferno, inferno_r, jet, jet_r, magma, magma_r, nipy_spectral, nipy_spectral_r, ocean, ocean_r, pink, pink_r, plasma, plasma_r, prism, prism_r, rainbow, rainbow_r, seismic, seismic_r, spring, spring_r, summer, summer_r, tab10, tab10_r, tab20, tab20_r, tab20b, tab20b_r, tab20c, tab20c_r, terrain, terrain_r, twilight, twilight_r, twilight_shifted, twilight_shifted_r, viridis, viridis_r, winter, winter_r

RuntimeError: cuda runtime error (48) : no kernel image is available for execution on the device

MY ENVIROMENT：

Name Version Build Channel

QUESTION：
(pytroch-yolov3) siat@siat-Precision-3640-Tower:~/work/object_detection/PyTorch-YOLOv3-kitti-master$ python detect.py
Config:
Namespace(batch_size=1, class_path='./data/kitti.names', conf_thres=0.8, config_path='./config/yolov3-kitti.cfg', image_folder='./data/samples/', img_size=416, n_cpu=4, nms_thres=0.4, use_cuda=True, weights_path='./weights/kitti.weights')
/home/siat/Downloads/ENTER/envs/pytroch-yolov3/lib/python3.5/site-packages/torch/nn/functional.py:52: UserWarning: size_average and reduce args will be deprecated, please use reduction='elementwise_mean' instead.
warnings.warn(warning.format(ret))
model path: ./weights/kitti.weights
using cuda model
data size : 20

Performing object detection:
THCudaCheck FAIL file=/opt/conda/conda-bld/pytorch_1535490206202/work/aten/src/THC/THCGeneral.cpp line=663 error=8 : invalid device function
Traceback (most recent call last):
File "detect.py", line 72, in
detections = model(input_imgs)
File "/home/siat/Downloads/ENTER/envs/pytroch-yolov3/lib/python3.5/site-packages/torch/nn/modules/module.py", line 477, in call
result = self.forward(*input, **kwargs)
File "/home/siat/work/object_detection/PyTorch-YOLOv3-kitti-master/models.py", line 256, in forward
x = module(x)
File "/home/siat/Downloads/ENTER/envs/pytroch-yolov3/lib/python3.5/site-packages/torch/nn/modules/module.py", line 477, in call
result = self.forward(*input, **kwargs)
File "/home/siat/Downloads/ENTER/envs/pytroch-yolov3/lib/python3.5/site-packages/torch/nn/modules/container.py", line 91, in forward
input = module(input)
File "/home/siat/Downloads/ENTER/envs/pytroch-yolov3/lib/python3.5/site-packages/torch/nn/modules/module.py", line 477, in call
result = self.forward(*input, **kwargs)
File "/home/siat/Downloads/ENTER/envs/pytroch-yolov3/lib/python3.5/site-packages/torch/nn/modules/activation.py", line 447, in forward
return F.leaky_relu(input, self.negative_slope, self.inplace)
File "/home/siat/Downloads/ENTER/envs/pytroch-yolov3/lib/python3.5/site-packages/torch/nn/functional.py", line 755, in leaky_relu
return torch._C._nn.leaky_relu(input, negative_slope)
RuntimeError: cuda runtime error (48) : no kernel image is available for execution on the device at /opt/conda/conda-bld/pytorch_1535490206202/work/aten/src/THCUNN/generic/LeakyReLU.cu:29

求问大佬

这个项目如果改变分类的数目，只识别car类型的话，需要进行什么样的改造？

how to use multi gpus?

I didn't find multiple gpu training code in repro, and I found that use nn.DataParallel also does not work.

packyan / pytorch-yolov3-kitti Goto Github PK

pytorch-yolov3-kitti's People

Contributors

Stargazers

Watchers

Forkers

pytorch-yolov3-kitti's Issues

Name Version Build Channel

Recommend Projects

Recommend Topics

Recommend Org