search

weiliu89 / caffe

Caffe: a fast open framework for deep learning.
http://caffe.berkeleyvision.org/
Other
4.77k stars 1.68k forks source link

About the result of ssd_detect.cpp and ssd_pascal_video.py #279

Open sxl2016 opened 7 years ago

sxl2016 commented 7 years ago

My ssd_detect.cpp's result is not right, it's result is different with ssd_pascal_video.py on the same video. Such as ILSVRC2015_train_00755001.mp4. What's may cause that? Is it need a mean_file? The commend I used is: cd ~/caffe-ssd ./build/examples/ssd/ssd_detect.bin \ ~/caffe-ssd/models/models/VGGNet/VOC0712/SSD_300x300/deploy.prototxt \ ~/caffe-ssd/models/models/VGGNet/VOC0712/SSD_300x300/VGG_VOC0712_SSD_300x300_iter_60000.caffemodel \ ~/Pictures/car.txt \ --file_type=video \ --confidence_threshold=0.3

weiliu89 commented 7 years ago

Can you be more specific what is the difference?

sxl2016 commented 7 years ago

![Uploading SSD_DETECT.png…]() ![Uploading SSD_DETECT_VIDEO.png…]()

@weiliu89 The result of ssd_detect.cpp is the first picture. And the second is ssd_detect_video.py. The results are different to the same video.

sxl2016 commented 7 years ago

ssd_detect

ssd_detect_video

weiliu89 commented 7 years ago

Are you sure you draw the boxes correctly?