Spaces:
Sleeping
Sleeping
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
---
|
| 2 |
title: ERA SESSION13
|
| 3 |
-
emoji:
|
| 4 |
colorFrom: indigo
|
| 5 |
colorTo: indigo
|
| 6 |
sdk: gradio
|
|
@@ -21,35 +21,6 @@ HF Link: https://huggingface.co/spaces/RaviNaik/ERA-SESSION13
|
|
| 21 |
4. **No obj accuracy: 97.991463%**
|
| 22 |
5. **Obj accuracy: 75.976616%**
|
| 23 |
6. **MAP: 0.4366795**
|
| 24 |
-
|
| 25 |
-
### Tasks:
|
| 26 |
-
1. :heavy_check_mark: Move the code to PytorchLightning
|
| 27 |
-
2. :heavy_check_mark: Train the model to reach such that all of these are true:
|
| 28 |
-
- Class accuracy is more than 75%
|
| 29 |
-
- No Obj accuracy of more than 95%
|
| 30 |
-
- Object Accuracy of more than 70% (assuming you had to reduce the kernel numbers, else 80/98/78)
|
| 31 |
-
- Ideally trained till 40 epochs
|
| 32 |
-
3. :heavy_check_mark: Add these training features:
|
| 33 |
-
- Add multi-resolution training - the code shared trains only on one resolution 416
|
| 34 |
-
- Add Implement Mosaic Augmentation only 75% of the times
|
| 35 |
-
- Train on float16
|
| 36 |
-
- GradCam must be implemented.
|
| 37 |
-
4. :heavy_check_mark: Things that are allowed due to HW constraints:
|
| 38 |
-
- Change of batch size
|
| 39 |
-
- Change of resolution
|
| 40 |
-
- Change of OCP parameters
|
| 41 |
-
5. :heavy_check_mark: Once done:
|
| 42 |
-
- Move the app to HuggingFace Spaces
|
| 43 |
-
- Allow custom upload of images
|
| 44 |
-
- Share some samples from the existing dataset
|
| 45 |
-
- Show the GradCAM output for the image that the user uploads as well as for the samples.
|
| 46 |
-
6. :heavy_check_mark: Mention things like:
|
| 47 |
-
- classes that your model support
|
| 48 |
-
- link to the actual model
|
| 49 |
-
7. :heavy_check_mark: Assignment:
|
| 50 |
-
- Share HuggingFace App Link
|
| 51 |
-
- Share LightningCode Link on Github
|
| 52 |
-
- Share notebook link (with logs) on GitHub
|
| 53 |
|
| 54 |
### Results
|
| 55 |
|
|
@@ -531,8 +502,4 @@ MAP: 0.43667954206466675
|
|
| 531 |
|
| 532 |
### GradCAM Representations
|
| 533 |
EigenCAM is used to generate CAM representation, since usal gradient based method wont work with detection models like Yolo, FRCNN etc.
|
| 534 |
-
|
| 535 |
-
|
| 536 |
-
|
| 537 |
-
|
| 538 |
-
|
|
|
|
| 1 |
---
|
| 2 |
title: ERA SESSION13
|
| 3 |
+
emoji: 🔥
|
| 4 |
colorFrom: indigo
|
| 5 |
colorTo: indigo
|
| 6 |
sdk: gradio
|
|
|
|
| 21 |
4. **No obj accuracy: 97.991463%**
|
| 22 |
5. **Obj accuracy: 75.976616%**
|
| 23 |
6. **MAP: 0.4366795**
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
### Results
|
| 26 |
|
|
|
|
| 502 |
|
| 503 |
### GradCAM Representations
|
| 504 |
EigenCAM is used to generate CAM representation, since usal gradient based method wont work with detection models like Yolo, FRCNN etc.
|
| 505 |
+
|
|
|
|
|
|
|
|
|
|
|
|