Comments (11)
I will need more details. What image size are you using? did you set the dimensions properly?
from fast_dense_feature_extraction.
Hi, thank you for the prompt response. I just use the sample script, nothing changed.
My pytorch version is 1.5, and cuda==10.1, win10
from fast_dense_feature_extraction.
Hi, 'singlePatch' mode works well. This issue only occurs in 'allPatches' mode
Total time for C_I: 0.013318886756896972sec
Total time for C_P per Patch without warm up: 0.0009350776672363281sec
------- Comparison between a base_net single patch evaluation and slim_net -------
difference between the base_net & slim_net - 0.0000000000000
from fast_dense_feature_extraction.
please check the size of cropped_patches
from fast_dense_feature_extraction.
slim_net_output_numpy.shape: (113, 960, 1280)
len(cropped_patches): 1228800
cropped_patches[0].shape: torch.Size([1, 1, 15, 15])
base_net_output_per_patch.shape: (113, 960, 1280)
cropped_batches.shape: torch.Size([10, 1, 15, 15])
y_base.shape: torch.Size([10, 113, 1, 1])
base_current_output.shape: (10, 113)
base_net_output_per_patch.shape: (113, 960, 1280)
index: 108153647
imH*imW: 1228800
base_net_output_per_patch.size: 138854400
from fast_dense_feature_extraction.
It is solved by changing the line 177
ci, yi, xi = np.unravel_index(index, (slim_net.outChans, imH, imW))
from fast_dense_feature_extraction.
Changed these line. You can pull again
index = np.argmax(sum(abs(base_net_output_per_patch - slim_net_output_numpy)))
max_diff = np.max(sum(abs(base_net_output_per_patch - slim_net_output_numpy)))
from fast_dense_feature_extraction.
Thanks a lot.
from fast_dense_feature_extraction.
Hi, I have one more question.
If my own network consists of layers without striding or pooling, How do I transform Cp to Ci.
Does the function multiConv help?
@erezposner
from fast_dense_feature_extraction.
Please refer to the paper https://arxiv.org/pdf/1805.03096.pdf.
"In this paper,
we present an approach to compute patch-based local feature descriptors efficiently in
presence of pooling and striding layers for whole images at once"
This means that without pooling or stride layers there isn't any difference between Cp and Ci
from fast_dense_feature_extraction.
I see, thanks.
from fast_dense_feature_extraction.
Related Issues (8)
- Can the transpose-reshape operation in unwarping layer be replaced by pixel shuffle? HOT 1
- This fast dense feature extraction method (FDFE) seems not equal to the common one in terms of different receptive field. HOT 2
- Example code for standard architectures like ResNet? HOT 1
- Please check typo HOT 2
- dimensions of the ouput HOT 1
- what can i do if my custom basenet feature extractor does not use any maxpool2d layers?
- What is the purpose of multiConv?
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from fast_dense_feature_extraction.