Comments (5)
@LouieYang It might be because of GNs in tower layers. Please remove all GN layers there and try again.
from fcos.
Thank you @tianzhi0549, the inference result doesn't change after I remove the GN layers.
I finally fix it by increasing INFERENCE_TH to 0.2 and it can overfit one image now. Still wondering why those boxes appear with 0.05 < scores < 0.2, any idea of it ?
from fcos.
@LouieYang I am not sure. It might be because we directly output the top 100 predictions after the post-processing, regardless of their scores.
from fcos.
@tianzhi0549 Seems like in inference, fcos filtered top 100 predictions according to their scores.
Thanks again for the nice code btw.
Closed (FCOS could correctly overfit one training sample)
from fcos.
Thank you @tianzhi0549, the inference result doesn't change after I remove the GN layers.
I finally fix it by increasing INFERENCE_TH to 0.2 and it can overfit one image now. Still wondering why those boxes appear with 0.05 < scores < 0.2, any idea of it ?
Hello, i also meet the same problem in mmdet version of fcos, i train the model with pytorch QAT.
Have you ever solve the problem?
And how do you train your model to overfit one image? I try to train one image with 100 epoch , it still cannot overfit.
from fcos.
Related Issues (20)
- Can I use it on CPU?
- Can we use this FCOS algorithm for counting the objects in images??
- May I ask you for a train log file?
- Which is the accuracy index in the paper?
- Error pip install git+https://github.com/tianzhi0549/FCOS.git HOT 1
- use my dataset to train FCOS net,why the loss is nan? HOT 2
- Can’t calculate the Params and FLOPs of Backbone
- Testing with test dataset while training
- ValueError: Unknown CUDA arch (8.6) or GPU not supported HOT 2
- weird time consuming of FCOS head when changing the backbone
- Problem with the last line of Testing-only installation HOT 1
- Deepcopy returns TypeError: 'int' object is not callable HOT 1
- assert len(proposal_losses) == 1 and proposal_losses["zero"] == 0 # loss_dict should be empty dict
- tools/train_net.py FAILED HOT 1
- Convert onnx
- Segmentation fault (core dumped)
- gempy to load csv data
- ImportError: cannot import name '_C'
- Why regression head wasn't normalized ?
- ValueError: num_samples should be a positive integer value, but got num_samples=0
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 fcos.