Update README.md
Browse files
README.md
CHANGED
|
@@ -1,9 +1,9 @@
|
|
|
|
|
| 1 |
# Car Bounding Box Detection — Custom CNN From Scratch
|
| 2 |
|
| 3 |
This repository contains a **custom Convolutional Neural Network (CNN)** trained **from scratch** for **car bounding box detection** on the **Stanford Cars Dataset**.
|
| 4 |
The model predicts bounding boxes in normalized format: `[x_center, y_center, width, height]`.
|
| 5 |
|
| 6 |
-
---
|
| 7 |
|
| 8 |
## Features
|
| 9 |
|
|
@@ -15,7 +15,6 @@ The model predicts bounding boxes in normalized format: `[x_center, y_center, wi
|
|
| 15 |
- Smooth L1 loss for bounding box regression
|
| 16 |
- Fully GPU-compatible training and inference
|
| 17 |
|
| 18 |
-
---
|
| 19 |
|
| 20 |
## Dataset
|
| 21 |
|
|
@@ -23,7 +22,6 @@ The model predicts bounding boxes in normalized format: `[x_center, y_center, wi
|
|
| 23 |
- **Annotations used:** Bounding boxes only
|
| 24 |
- Images resized to **416×416 pixels**
|
| 25 |
|
| 26 |
-
---
|
| 27 |
|
| 28 |
## Model Architecture
|
| 29 |
|
|
@@ -33,7 +31,6 @@ The model predicts bounding boxes in normalized format: `[x_center, y_center, wi
|
|
| 33 |
- Sigmoid output to produce normalized coordinates
|
| 34 |
- Output format: `[x_center, y_center, width, height]`
|
| 35 |
|
| 36 |
-
---
|
| 37 |
|
| 38 |
## Training
|
| 39 |
|
|
@@ -43,7 +40,6 @@ The model predicts bounding boxes in normalized format: `[x_center, y_center, wi
|
|
| 43 |
- **Scheduler:** Cosine annealing LR
|
| 44 |
- Training monitored with best validation IoU checkpointing
|
| 45 |
|
| 46 |
-
---
|
| 47 |
|
| 48 |
## Inference
|
| 49 |
|
|
@@ -55,9 +51,11 @@ The model predicts bounding boxes in normalized format: `[x_center, y_center, wi
|
|
| 55 |
|
| 56 |
## Example
|
| 57 |
|
|
|
|
|
|
|
| 58 |
|
| 59 |
|
| 60 |
-
##
|
| 61 |
|
| 62 |
If you use this model, please cite:
|
| 63 |
|
|
@@ -72,6 +70,6 @@ If you use this model, please cite:
|
|
| 72 |
|
| 73 |
```
|
| 74 |
|
| 75 |
-
##
|
| 76 |
|
| 77 |
license : MIT
|
|
|
|
| 1 |
+
|
| 2 |
# Car Bounding Box Detection — Custom CNN From Scratch
|
| 3 |
|
| 4 |
This repository contains a **custom Convolutional Neural Network (CNN)** trained **from scratch** for **car bounding box detection** on the **Stanford Cars Dataset**.
|
| 5 |
The model predicts bounding boxes in normalized format: `[x_center, y_center, width, height]`.
|
| 6 |
|
|
|
|
| 7 |
|
| 8 |
## Features
|
| 9 |
|
|
|
|
| 15 |
- Smooth L1 loss for bounding box regression
|
| 16 |
- Fully GPU-compatible training and inference
|
| 17 |
|
|
|
|
| 18 |
|
| 19 |
## Dataset
|
| 20 |
|
|
|
|
| 22 |
- **Annotations used:** Bounding boxes only
|
| 23 |
- Images resized to **416×416 pixels**
|
| 24 |
|
|
|
|
| 25 |
|
| 26 |
## Model Architecture
|
| 27 |
|
|
|
|
| 31 |
- Sigmoid output to produce normalized coordinates
|
| 32 |
- Output format: `[x_center, y_center, width, height]`
|
| 33 |
|
|
|
|
| 34 |
|
| 35 |
## Training
|
| 36 |
|
|
|
|
| 40 |
- **Scheduler:** Cosine annealing LR
|
| 41 |
- Training monitored with best validation IoU checkpointing
|
| 42 |
|
|
|
|
| 43 |
|
| 44 |
## Inference
|
| 45 |
|
|
|
|
| 51 |
|
| 52 |
## Example
|
| 53 |
|
| 54 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/67bc31088cf27f32cbcf927f/h286qIktC-H5CkxuO-YvH.jpeg" width="300"/>
|
| 55 |
+
|
| 56 |
|
| 57 |
|
| 58 |
+
## Citation
|
| 59 |
|
| 60 |
If you use this model, please cite:
|
| 61 |
|
|
|
|
| 70 |
|
| 71 |
```
|
| 72 |
|
| 73 |
+
## License
|
| 74 |
|
| 75 |
license : MIT
|