Comments (5)
Hi, @RunningLeon Could you please handle this issue?
from mmdeploy.
@tehkillerbee Hi, you could create your custom deploy_cfg file by changing keep_top_k=1000
. And max_output_boxes_per_class
is controlling max number of detection results for each class.
A simple way is to inherit and overwrite by creating a new deploy config file:
# in mmdeploy/configs/mmdet/detection/new_config.py
_base_ = ['../_base_/base_tensorrt_dynamic-320x320-1344x1344.py']
# got everything from base config
# overwriting parameters you want to change
codebase_config = dict(
post_processing=dict(
max_output_boxes_per_class=200,
keep_top_k=1000 # update to 1000
))
from mmdeploy.
@RunningLeon Ok, that is a good solution then - similar to how the mmdet configs can inherit from eachother and how values can be overwritten.
FYI, I am working on a model, where only a single class is detected, but with a high number of detections per image (eg. 1000) so this was why I set the max_output_boxes_per_class
higher. Previously, I did not change keep_top_k
, but I still got more than 100 results.
from mmdeploy.
@RunningLeon Ok, that is a good solution then - similar to how the mmdet configs can inherit from eachother and how values can be overwritten.
FYI, I am working on a model, where only a single class is detected, but with a high number of detections per image (eg. 1000) so this was why I set the
max_output_boxes_per_class
higher. Previously, I did not changekeep_top_k
, but I still got more than 100 results.
@tehkillerbee Yes, you're right.
That's because parameters are set based on original model config at first. If nothing found in original model config, then it would be set based on deploy config.
You could take a look at here
from mmdeploy.
@RunningLeon Thanks for the help, that makes sense.
Closing this issue, as it is not a MMdeploy bug
from mmdeploy.
Related Issues (20)
- [Bug]Build MMDeploy get wrong! HOT 2
- [Bug] Error Exporting RetinaNet for Single Class Case with CrossEntropyLoss in MMDeploy HOT 1
- [Bug] RTMDet : Error when deploying onnx/engine model using tensorRT backend configs HOT 3
- [Bug] Cannot use Tritonserver with MMdeploy custom operations
- ROIAlign plugin was not found in the plugin registry!
- [Bug] Build mmdeploy sdk for tensorrt err
- [Bug] Unable to reproduce export to CoreML in lib HOT 1
- [Docs] A100算力加持!书生大模型实战营第3期全面升级,趣味闯关模式等你开启
- [Bug] Warning: The shape inference of mmdeploy::Mark type is missing HOT 1
- [Bug] mmdeploy模型部署-pytorch转换成onnx
- [Bug] mmocr - dbnet train model test result is difference with mmdeploy GPU inference result HOT 1
- [Bug] the accuracy of pose estimation has deteriorated.
- [Bug] Plugin load error in board infer HOT 3
- [Feature] Class agnostic NMS for object detector in torch2onnx
- [Bug] It works for ONNX but it's can works for NCNN. HOT 2
- [Bug] Bounding box coordinates returned by the pose tracker API are not scaled correctly HOT 1
- RTMDET Tensorrt c++ detector apply images failed。[Bug] HOT 1
- [Bug] DDQDETR export to onnx error
- [Bug] C++ SDK batch inference not working? HOT 1
- [Feature] Is mmdeploy ncnn support gpu on windows
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 mmdeploy.