- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
I transfered my caffe model to Movidius' model, and encountered strange problems that the output value is the same on any image.
I use the mvNCCheck to debug, discover the value became the same after bn_conv1 layer.
Here is the mvNCCheck Result: https://drive.google.com/open?id=1jM6k3ki8ZI9MRi88bCYkpvvRN74NDjwWf669LvWku_I
$ mvNCCheck -w run1_iter_38880.caffemodel deploy.prototxt -s 12 -on bn_conv1
Result: (32, 336, 336)
1) 3612671 0.0
2) 1204219 0.0
3) 1204230 0.0
4) 1204229 0.0
5) 1204228 0.0
Expected: (32, 336, 336)
1) 1655946 -0.36279
2) 2072916 -0.36304
3) 2085841 -0.36523
4) 2117228 -0.3667
5) 2066852 -0.36743
Where's wrong and why?
Here is my caffemodel and part of deploy.protxt.
https://drive.google.com/drive/folders/1Dtvo4lvTONTIrWkrrwiGtTw7v1yO4qu9
Please assist me, thank you.
Link Copied
1 Reply
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
@miily I'll take a look at it and see what I can find. It is strange that the Caffe results are also showing the similar probabilities across the top 5 results (negative results at that). The class numbers are pretty large, just wondering what kind of network this is. Thanks.
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page