this repository is designed for training Faster RCNN and YOLOv3 models on PASCAL-VOC dataset using mmdetection codebase.
please refer to https://github.com/open-mmlab/mmdetection
download this repository
download pascal-voc dataset and place it to mmdet_VOC/data run the given scripts to switch the VOC data format to COCO data format
cd mmdet_VOC
# remember to modify path variables in the following python file
python voc_to_coco.py # this will create json files in coco format
python split.py # this will reorganize images to coco format
now we have a coco format voc dataset for further training and evaluation
# logfiles will be saved to mmdet_VOC/faster_rcnn
# if needed, modify hyper parameters in the config file
python tools/train.py configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py --gpus 1 --work-dir faster_rcnn # this command only supports single GPU training
losses and evaluation results will be strored at mmdet_VOC/faster_rcnn/*.log.json, the model will save latest checkpoint after every epoch and evaluate by then
you can use tensorboard to check the loss curves during the training process
# create logfile
python create_tensorboard.py # remenber to modify base_dir, train/eval/test, step
tensorboard --logdir=faster_rcnn/log_file/train
test the performance of the model
python tools/test.py configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py faster_rcnn/latest.pth --work-dir faster_rcnn --eval bbox`
inference images
python inference.py # remember to modify config_file, checkpoint_file and imgs
# if u don't offer out_file when using show_result_pyplot(), inference img will be shown to screen
commands are similar to the previous section, remenber to use configs/yolo/yolov3_d53_mstrain-608_273e_coco.py as your config file, and modify offered scripts for visualization and inference if u need.
you can download my pretrained model(PASCAL2007 dataset) from Baidu NetDisk https://pan.baidu.com/s/15K6iNJncWQVtIncFi7HMlg, access code is s7c7.
given codes only support training and inference on PASCAL-VOC dataset, if u want to use your own dataset, remember to organize them in COCO format or just organize them in PASCAL-VOC format and switch to COCO format using given voc_to_coco.py script. and then please make sure that u change the dataset classes written in the config file to avoid error.
if you want to visualize the proposals given by the RPN module in faster rcnn, u can check ~/anaconda/envs/openmmlab/python3.7/site-packages/mmdet/detectors/two_stage.py and add the following code before the return line of function simple_test()
#######################################
# for visualizing the proposals given by RPN
import numpy as np
import matplotlib.pyplot as plt
bboxes = proposal_list[0].cpu().numpy()
bboxes = bboxes[bboxes[:, 4] > 0.7][:50, :]
img_meta = img_metas[0]
bboxes[:, :4] = bboxes[:, :4] / img_meta['scale_factor'][0]
img = mmcv.imread(img_meta['filename']).astype(np.uint8)
img = img.copy()
img = mmcv.bgr2rgb(img)
width, height = img.shape[1], img.shape[0]
img = np.ascontiguousarray(img)
win_name = 'RPN_proposals'
fig = plt.figure(win_name, frameon=False)
plt.title(win_name)
canvas = fig.canvas
dpi = fig.get_dpi()
# add a small EPS to avoid precision lost due to matplotlib's truncation
# (https://github.com/matplotlib/matplotlib/issues/15363)
EPS = 1e-2
fig.set_size_inches((width + EPS) / dpi, (height + EPS) / dpi)
# remove white edges by set subplot margin
plt.subplots_adjust(left=0, right=1, bottom=0, top=1)
ax = plt.gca()
ax.axis('off')
# max_label = int(max(labels) if len(labels) > 0 else 0)
# text_palette = palette_val(get_palette(text_color, max_label + 1))
# text_colors = [text_palette[label] for label in labels]
from mmdet.core.visualization.palette import get_palette, palette_val
from mmdet.core.visualization.image import draw_bboxes
bbox_color = (72, 101, 241)
thickness = 0.8
num_bboxes = 0
if bboxes is not None:
num_bboxes = bboxes.shape[0]
bbox_palette = palette_val(get_palette(bbox_color, 1))
colors = bbox_palette*num_bboxes
draw_bboxes(ax, bboxes, colors, alpha=0.8, thickness=thickness)
horizontal_alignment = 'left'
positions = bboxes[:, :2].astype(np.int32) + thickness
plt.imshow(img)
stream, _ = canvas.print_to_buffer()
buffer = np.frombuffer(stream, dtype='uint8')
img_rgba = buffer.reshape(height, width, 4)
rgb, alpha = np.split(img_rgba, [3], axis=2)
img = rgb.astype('uint8')
img = mmcv.rgb2bgr(img)
import os
mmcv.imwrite(img, os.path.join('faster_rcnn/rpn_proposals', os.path.basename(img_meta['filename'])))
################################################
and then run inference.py to save the proposals of given image
if you have any other problems, feel free to let me know
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent