NanoDet-Plus ONNX models
Four pretrained NanoDet-Plus COCO object detection models from the official NanoDet project.
| Folder | Input size |
|---|---|
plus-m-320/ |
320 × 320 |
plus-m-416/ |
416 × 416 |
plus-m-1.5x-320/ |
320 × 320 |
plus-m-1.5x-416/ |
416 × 416 |
Each folder contains an ONNX model and a matching labels.txt.
Labels are one per line, in zero-based COCO class order.
Examples: person = 0, car = 2, motorcycle = 3.
These are pretrained models published by the NanoDet authors. This repository repackages the original ONNX files with their labels; it does not claim to have trained or fine-tuned them.
Source
- Project: https://github.com/RangiLyu/nanodet
- Original releases: https://github.com/RangiLyu/nanodet/releases
- License: https://github.com/RangiLyu/nanodet/blob/main/LICENSE
Inference
The ONNX output contains class scores and box distributions. NanoDet-specific box decoding and non-maximum suppression are required; the raw ONNX output is not a final list of bounding boxes.