| language: en | |
| license: mit | |
| tags: | |
| - yolov8 | |
| - object-detection | |
| - waste-classification | |
| library_name: onnx | |
| pipeline_tag: object-detection | |
| # trashscan8m | |
| YOLOv8 mobile_model_package_yolov8m_v2 ONNX model for waste classification. | |
| ## Classes | |
| | Index | Class | | |
| |-------|-------| | |
| | 0 | paper | | |
| | 1 | plastic | | |
| | 2 | metal | | |
| | 3 | organic | | |
| | 4 | other | | |
| ## Usage | |
| ```python | |
| import onnxruntime | |
| import numpy as np | |
| from PIL import Image | |
| session = onnxruntime.InferenceSession("best.onnx") | |
| input_name = session.get_inputs()[0].name | |
| # Input: [1, 3, 640, 640] normalized float32 | |
| # Output: [1, 9, 8400] = [cx, cy, w, h, p0, p1, p2, p3, p4] | |
| ``` | |
| ## Model Specs | |
| - Input size: 640x640 | |
| - Format: ONNX | |
| - Confidence threshold: 0.25 | |
| - IoU threshold: 0.5 | |