| --- |
| license: mit |
| tags: |
| - image-classification |
| - object-detection |
| - waste-classification |
| - recycling |
| - mobilenet |
| - yolov8 |
| - tensorflow |
| - pytorch |
| --- |
| |
| # Smart Waste Classification Models |
|
|
| Is repository mein 2 trained models hain jo waste (kachra) images ko classify karne ke liye use hote hain. Dono models ek Flask/Gradio app mein integrate kiye gaye hain jahan user apni marzi se koi bhi ek model select kar sakta hai. |
|
|
| ## Models |
|
|
| ### 1. `best_mobilenet.keras` — MobileNetV2 (Image Classification) |
| |
| - **Task:** Whole-image classification (5 classes) |
| - **Framework:** TensorFlow / Keras |
| - **Input size:** 224x224 RGB image |
| - **Classes:** |
| - cardboard |
| - glass |
| - metal |
| - paper |
| - plastic |
| - trash |
| |
| Ye model poori image ko dekh kar batata hai ke image mein sabse zyada kis waste category ka material hai. Har class ke liye ek confidence score bhi milta hai. |
| |
| ### 2. `bestyolomodel.pt` — YOLOv8 (Object Detection) |
| |
| - **Task:** Object detection (3 classes) |
| - **Framework:** Ultralytics YOLOv8 / PyTorch |
| - **Classes:** 3 waste categories (bounding-box ke sath localization) |
| |
| Ye model image ke andar waste object ko detect karta hai aur uske around bounding box + class + confidence deta hai. MobileNet ke muqable ye batata hai ke object **kahan** hai, sirf ye nahi ke image mein kya hai. |
| |
| ## Recyclability Mapping |
| |
| Dono models ke output ko is mapping ke zariye Recyclable / Non-Recyclable mein convert kiya jata hai: |
| |
| | Class | Status | |
| |------------|----------------| |
| | cardboard | Recyclable | |
| | glass | Recyclable | |
| | metal | Recyclable | |
| | paper | Recyclable | |
| | plastic | Recyclable | |
| | trash | Non-Recyclable | |
| |
| ## Usage |
| |
| ```python |
| from huggingface_hub import hf_hub_download |
| from tensorflow.keras.models import load_model |
| from ultralytics import YOLO |
| |
| REPO_ID = "SabaTariq510/waste-classification-models" |
|
|
| # MobileNetV2 |
| mobilenet_path = hf_hub_download(repo_id=REPO_ID, filename="best_mobilenet.keras") |
| mobilenet_model = load_model(mobilenet_path) |
| |
| # YOLOv8 |
| yolo_path = hf_hub_download(repo_id=REPO_ID, filename="bestyolomodel.pt") |
| yolo_model = YOLO(yolo_path, task="detect") |
| ``` |
| |
| ## App |
| |
| Ye models ek Gradio app mein deploy kiye gaye hain jahan user image upload kar ke MobileNetV2 ya YOLOv8 mein se koi bhi model select kar sakta hai prediction ke liye. |
| |
