|
--- |
|
license: apache-2.0 |
|
metrics: |
|
- precision |
|
base_model: |
|
- Ultralytics/YOLO11s |
|
tags: |
|
- object_detection |
|
model-index: |
|
- name: Object Detection Model |
|
results: |
|
- task: |
|
type: object-detection |
|
dataset: |
|
name: Custom Object Dataset |
|
type: object-detection |
|
metrics: |
|
- name: Box(P) |
|
type: precision |
|
value: 0.904 |
|
- name: R |
|
type: recall |
|
value: 0.87 |
|
- name: mAP50 |
|
type: mAP |
|
value: 0.918 |
|
- name: mAP50-95 |
|
type: mAP |
|
value: 0.671 |
|
details: |
|
- class: |
|
name: all |
|
images: 92 |
|
instances: 2568 |
|
metrics: |
|
- Box(P): 0.904 |
|
- R: 0.87 |
|
- mAP50: 0.918 |
|
- mAP50-95: 0.671 |
|
- class: |
|
name: range |
|
images: 82 |
|
instances: 82 |
|
metrics: |
|
- Box(P): 0.928 |
|
- R: 0.938 |
|
- mAP50: 0.957 |
|
- mAP50-95: 0.701 |
|
- class: |
|
name: entry_door |
|
images: 92 |
|
instances: 821 |
|
metrics: |
|
- Box(P): 0.941 |
|
- R: 0.944 |
|
- mAP50: 0.966 |
|
- mAP50-95: 0.704 |
|
- class: |
|
name: kitchen_sink |
|
images: 80 |
|
instances: 91 |
|
metrics: |
|
- Box(P): 0.863 |
|
- R: 0.828 |
|
- mAP50: 0.917 |
|
- mAP50-95: 0.662 |
|
- class: |
|
name: bathroom_sink |
|
images: 89 |
|
instances: 240 |
|
metrics: |
|
- Box(P): 0.909 |
|
- R: 0.85 |
|
- mAP50: 0.929 |
|
- mAP50-95: 0.64 |
|
- class: |
|
name: toilet |
|
images: 90 |
|
instances: 188 |
|
metrics: |
|
- Box(P): 0.927 |
|
- R: 0.904 |
|
- mAP50: 0.96 |
|
- mAP50-95: 0.667 |
|
- class: |
|
name: double_folding_door |
|
images: 19 |
|
instances: 37 |
|
metrics: |
|
- Box(P): 0.867 |
|
- R: 0.702 |
|
- mAP50: 0.828 |
|
- mAP50-95: 0.594 |
|
- class: |
|
name: window |
|
images: 88 |
|
instances: 669 |
|
metrics: |
|
- Box(P): 0.871 |
|
- R: 0.9 |
|
- mAP50: 0.905 |
|
- mAP50-95: 0.582 |
|
- class: |
|
name: shower |
|
images: 61 |
|
instances: 70 |
|
metrics: |
|
- Box(P): 0.907 |
|
- R: 0.957 |
|
- mAP50: 0.947 |
|
- mAP50-95: 0.778 |
|
- class: |
|
name: bathtub |
|
images: 71 |
|
instances: 103 |
|
metrics: |
|
- Box(P): 0.947 |
|
- R: 0.874 |
|
- mAP50: 0.933 |
|
- mAP50-95: 0.793 |
|
- class: |
|
name: single_folding_door |
|
images: 55 |
|
instances: 144 |
|
metrics: |
|
- Box(P): 0.877 |
|
- R: 0.839 |
|
- mAP50: 0.9 |
|
- mAP50-95: 0.647 |
|
- class: |
|
name: dishwasher |
|
images: 49 |
|
instances: 54 |
|
metrics: |
|
- Box(P): 0.912 |
|
- R: 0.833 |
|
- mAP50: 0.863 |
|
- mAP50-95: 0.568 |
|
- class: |
|
name: refrigerator |
|
images: 66 |
|
instances: 69 |
|
metrics: |
|
- Box(P): 0.901 |
|
- R: 0.87 |
|
- mAP50: 0.916 |
|
- mAP50-95: 0.712 |
|
source: |
|
name: Custom Object Detection Results |
|
url: https://example.com/custom-object-detection-results |
|
--- |
|
|
|
# YOLO11s Auto-CAD Detection Model Card |
|
|
|
## Model Overview |
|
The YOLO11s Auto-CAD detection model is a computer vision model trained to detect various objects related to AutoCAD layouts, such as kitchen sinks, toilets, windows, and other fixtures. This model is based on the YOLO11s architecture and fine-tuned for Auto-CAD-specific object detection tasks. The model is optimized for real-time inference on both GPU and CPU platforms. |
|
|
|
### Version Information |
|
- **Model Version**: YOLO11 |
|
- **Ultralytics Version**: 8.3.8 |
|
- **Python Version**: 3.9.7 |
|
- **Torch Version**: 2.3.1+cu118 |
|
- **CUDA Version**: 11.8 (for GPU use) |
|
- **Hardware**: NVIDIA GeForce RTX 4060 Laptop GPU, AMD Ryzen 7 7745HX (for CPU evaluation) |
|
- **Model Architecture**: YOLO (You Only Look Once) v5 (fused) |
|
|
|
## Model Details |
|
|
|
- **Layers**: 238 |
|
- **Parameters**: 9,417,444 |
|
- **GFLOPs**: 21.3 |
|
- **Training Type**: Supervised learning |
|
- **Target Platform**: GPUs (NVIDIA RTX series) and CPUs (AMD Ryzen) |
|
|
|
### Supported Tasks |
|
- Object detection on AutoCAD layouts |
|
- Classification and localization of various CAD objects |
|
|
|
## Dataset |
|
The model was trained on a dataset of AutoCAD object instances, including: |
|
- **Range** |
|
- **Entry Door** |
|
- **Kitchen Sink** |
|
- **Bathroom Sink** |
|
- **Toilet** |
|
- **Double Folding Door** |
|
- **Window** |
|
- **Shower** |
|
- **Bathtub** |
|
- **Single Folding Door** |
|
- **Dishwasher** |
|
- **Refrigerator** |
|
|
|
Each object class has a varying number of images and instances in the training set. |
|
|
|
## Performance Metrics |
|
|
|
| Class | Precision | Recall | mAP50 | mAP50-95 | |
|
|--------------------|-----------|--------|-------|----------| |
|
| all | 0.904 | 0.87 | 0.918 | 0.671 | |
|
| range | 0.928 | 0.938 | 0.957 | 0.701 | |
|
| entry_door | 0.941 | 0.944 | 0.966 | 0.704 | |
|
| kitchen_sink | 0.863 | 0.828 | 0.917 | 0.662 | |
|
| bathroom_sink | 0.909 | 0.85 | 0.929 | 0.64 | |
|
| toilet | 0.927 | 0.904 | 0.96 | 0.667 | |
|
| double_folding_door| 0.867 | 0.702 | 0.828 | 0.594 | |
|
| window | 0.871 | 0.9 | 0.905 | 0.582 | |
|
| shower | 0.907 | 0.957 | 0.947 | 0.778 | |
|
| bathtub | 0.947 | 0.874 | 0.933 | 0.793 | |
|
| single_folding_door| 0.877 | 0.839 | 0.9 | 0.647 | |
|
| dishwasher | 0.912 | 0.833 | 0.863 | 0.568 | |
|
| refrigerator | 0.901 | 0.87 | 0.916 | 0.712 | |
|
|
|
## Inference Speed |
|
- **Preprocess Time**: 0.2ms per image |
|
- **Inference Time**: 13.4ms per image |
|
- **Postprocess Time**: 1.5ms per image |
|
|
|
These times may vary depending on the hardware platform and the number of objects in the input image. |
|
|
|
## Usage |
|
|
|
### Requirements |
|
- Python 3.9.7 |
|
- PyTorch 2.3.1+cu118 |
|
- CUDA-enabled GPU (optional but recommended for faster inference) |
|
- Ultralytics YOLO package 8.3.8 |
|
|
|
### Installation |
|
To use the model, install the necessary dependencies: |
|
|
|
```bash |
|
pip install torch==2.3.1+cu118 |
|
pip install ultralytics |
|
``` |
|
|
|
## Demo |
|
You can try the model in action through this interactive demo: |
|
[**YOLO11 Auto-CAD Detection Demo**](https://huggingface.co/spaces/sabaridsnfuji/Symbol_SpottingonDigital_Architectural_FloorPlans_Using_objectdetection) |
|
|
|
## Notes |
|
- The model supports various object categories in AutoCAD drawings, such as doors, sinks, bathtubs, etc. |
|
- Performance metrics like mAP50 and mAP50-95 indicate the accuracy of detection and classification across multiple object categories. |
|
- This model is optimized for both GPU and CPU, with higher performance on GPUs. The model can be used for real-time detection applications requiring accurate localization of AutoCAD objects. |
|
|
|
--- |
|
|
