Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- object-detection
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
## Dataset
|
| 7 |
+
|
| 8 |
+
The dataset was referenced in the Smartathon competition.It's consist of 7874 images annontated with 11 classes:
|
| 9 |
+
|
| 10 |
+
* GARBAGE 8597
|
| 11 |
+
* CONSTRUCTION_ROAD 2730
|
| 12 |
+
* POTHOLES 2625
|
| 13 |
+
* CLUTTER_SIDEWALK 2253
|
| 14 |
+
* BAD_BILLBOARD 1555
|
| 15 |
+
* GRAFFITI 1124
|
| 16 |
+
* SAND_ON_ROAD 748
|
| 17 |
+
* UNKEPT_FACADE 127
|
| 18 |
+
* FADED_SIGNAGE 107
|
| 19 |
+
* BROKEN_SIGNAGE 83
|
| 20 |
+
* BAD_STREETLIGHT 1
|
| 21 |
+
|
| 22 |
+
The dataset highly imbalanced and contain some humman errors.
|
| 23 |
+
|
| 24 |
+
## Our SEE Team Solution
|
| 25 |
+
|
| 26 |
+
1. Convert from Pascal VOC to YOLO format
|
| 27 |
+
2. Model Hyperparamter tuning
|
| 28 |
+
3. Train the data on Yolov7
|
| 29 |
+
4. Evaluate the model
|
| 30 |
+
5. Expalin Different techniques to Automation of Data Annotation
|
| 31 |
+
|
| 32 |
+
For our solution detials: [notebook](https://colab.research.google.com/drive/1mo3HxJrg8wDGp_FhkB_0qAs41XvQ3hjR?usp=sharing)
|
| 33 |
+
|
| 34 |
+
### How to use
|
| 35 |
+
|
| 36 |
+
1. You can just download file weights from the files section
|
| 37 |
+
2. clone yolov7 repo ```!git clone https://github.com/WongKinYiu/yolov7```
|
| 38 |
+
3. ensure your current working directory is yolov7 then run ```! pip install -r requirements.txt```
|
| 39 |
+
4. then run the detector script ```! python detect.py --weights " model.pt path" --img 736 --conf 0.27 --source "testing image path" --save-txt```
|