Replies: 4 comments 2 replies
-
Any plan for the DLA backbone? If needed, I can try to implement it with fcos 😄. |
Beta Was this translation helpful? Give feedback.
-
Hello! Thanks a lot for your hard work! I have my own dataset for 3 classes in COCO format. I trained the model on this data. Is there currently a way to use the trained model to get metrics on the test data (test data is labeled)? I tried to run the script https://github.com/open-mmlab/mmdetection/blob/master/tools/analysis_tools/eval_metric.py but an error occurs, I think it is because it processes only the original COCO dataset for 80 classes.
But I managed to run this script https://github.com/open-mmlab/mmdetection/blob/master/tools/analysis_tools/analyze_results.py on my 3 classes by setting the parameter I've read all the documentation but haven't found a way to get quality scores on user test data. |
Beta Was this translation helpful? Give feedback.
-
great job, but i would like to see the support for tflite so it can be easily deployed to android devices. I can contribute to the code for such conversion, 'cus i went through quite a few issues converting the .pth model to tflite try to see if i can use the model on android . |
Beta Was this translation helpful? Give feedback.
-
Inference on video streams. Just the same YOLOv8 does. |
Beta Was this translation helpful? Give feedback.
-
Hi, the developers
We are glad to release MMDet 3.x since Sept 1st, 2022, with so many exciting changes providing stronger and more general benchmarks and cutting edges with easier usage for object detection. You can find the details in the initial release.
As MMDet 3.x is brand new, we are still completing its functionalities and documentation and we welcome any kind of contribution including issues/PRs/discussions. Moreover, we want to hear more from you so that MMDet 3.x could suit your projects of any purpose.
Therefore, we keep this discussion open to collect suggestions/ideas from users and hear your voice. Our monthly release plan will be updated in different milestones, and we will open discussion for each future release of MMDet 3.x, which will also be listed here. You can also leave Chinese message in #8779 if you prefer.
In this discussion, you can either:
Beta Was this translation helpful? Give feedback.
All reactions