pauhidalgoo commited on
Commit
b96bc27
·
verified ·
1 Parent(s): b42590d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +66 -1
README.md CHANGED
@@ -2,4 +2,69 @@
2
  base_model:
3
  - Ultralytics/YOLOv8
4
  pipeline_tag: object-detection
5
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
  base_model:
3
  - Ultralytics/YOLOv8
4
  pipeline_tag: object-detection
5
+ ---
6
+
7
+ # Backbone Finetuned YOLO Models for Oriented Object Detection
8
+
9
+ ## Overview
10
+ This repository contains finetuned YOLO models using a specific backbone for two sizes: Nano and Small. These models were trained and evaluated on the DIOR dataset, showcasing their performance in satellite object detection tasks. The finetuning process emphasized optimizing both training time and detection metrics.
11
+
12
+ This YOLO models were finetuned to perform [Oriented Bounding Box detection](https://docs.ultralytics.com/es/tasks/obb/#visual-samples).
13
+
14
+ ## Usage
15
+ The models can be used for oriented object detection tasks and are compatible with the [YOLO framework](https://docs.ultralytics.com). Detailed instructions for downloading and using these models are provided in the `README`.
16
+
17
+ ## Dataset
18
+ The models were trained on the [DIOR Dataset](https://paperswithcode.com/dataset/dior), which is tailored for detecting elements in satellite images. Evaluation metrics include mAP50 and mAP50-95 for a comprehensive assessment of detection accuracy.
19
+
20
+ ---
21
+
22
+ ## Model Performance and Comparison Table
23
+
24
+ | Model | Training Time (s) | mAP50 | mAP50-95 |
25
+ |------------------|-------------------|--------|----------|
26
+ | Backbone Nano | 7861.18 | 0.7552 | 0.5905 |
27
+ | Backbone Small | 7719.13 | 0.7824 | 0.6219 |
28
+
29
+ ---
30
+
31
+ ## Key Observations
32
+ - **Backbone Nano**: Achieves a respectable mAP50-95 of 0.5905, with moderate training time.
33
+ - **Backbone Small**: Outperforms the Nano model in both mAP50 and mAP50-95, while requiring slightly less training time.
34
+
35
+ ---
36
+
37
+ ## Examples
38
+
39
+
40
+ ![image/jpeg](https://cdn-uploads.huggingface.co/production/uploads/66605a873adae5fadc3ec6cf/aXRUjndIBsSgzlUwYF03X.jpeg)
41
+
42
+ ![image/png](https://huggingface.co/pauhidalgoo/yolov8-DIOR/resolve/main/DIOR_yolov8s_confusion_matrix_normalized.png)
43
+
44
+ ![image/png](https://huggingface.co/pauhidalgoo/yolov8-DIOR/resolve/main/DIOR_yolov8s_F1_curve.png)
45
+
46
+ ---
47
+
48
+ ## How to Use the Models
49
+ 1. Clone the repository.
50
+ 2. Install the ultralytics library (pip install ultralytics)
51
+ 3. Load the model size of your choice.
52
+
53
+ ```python
54
+ from ultralytics import YOLO
55
+
56
+ # Load a finetuned YOLO model
57
+ model = YOLO('path-to-model.pt')
58
+
59
+ # Perform inference
60
+ results = model('path-to-image.jpg')
61
+ results.show()
62
+ ```
63
+
64
+ ---
65
+
66
+ If your goal is to use it in matlab, you need to:
67
+ 1. Clone the repository.
68
+ 2. Clone the [Matlab YOLOv8](https://github.com/matlab-deep-learning/Pretrained-YOLOv8-Network-For-Object-Detection) repository.
69
+ 3. Use the [convert_to_onnx.py](https://huggingface.co/pauhidalgoo/yolov8-indoor/blob/main/convert_to_onnx.py)
70
+ 4. Load the model in Matalb like shown in [3_YOLO_matlab.mlx](https://huggingface.co/pauhidalgoo/yolov8-indoor/blob/main/3_YOLO_matlab.mlx)