Model Card for universal-logo-detector

Wandb tracking run

This object detection model was fine-tuned using the Ultralytics YOLO library.

Model Details

Model Description

• Developed by: Open Food Facts
• Model type: object detection
• License: agpl-3.0
• Finetuned from model [optional]: yolov8s.pt

Training Details

Training Data

The model was fine-tuned using the following dataset: openfoodfacts/universal-logo-detector (revision: v1.4).

Training Procedure

Dependency versions:

• ultralytics: 8.3.223
• pytorch: 2.9.0+cu128

Training Hyperparameters

• Epochs: 300
• Batch size: 64
• Image size: 640

Evaluation

The following evaluation metrics were obtained after training the model:

• metrics/precision(B): 0.8570181840729786

• metrics/recall(B): 0.7192719092719092

• metrics/mAP50(B): 0.8371955004759525

• metrics/mAP50-95(B): 0.6379071854832687

• fitness: 0.6379071854832687

Evaluation on exported models

The model was also evaluated after exporting to ONNX and TensorRT formats. The following metrics were obtained:

ONNX export

• metrics/precision(B): 0.8934595473774011

• metrics/recall(B): 0.68

• metrics/mAP50(B): 0.8107887350032505

• metrics/mAP50-95(B): 0.6517839531367096

• fitness: 0.6517839531367096

TENSORRT export

• metrics/precision(B): 0.893459487354923

• metrics/recall(B): 0.68

• metrics/mAP50(B): 0.8107887350032505

• metrics/mAP50-95(B): 0.6517839531367096

• fitness: 0.6517839531367096

Files

Most files stored on the repo are standard files created during training with the Ultralytics YOLO library.

What was added:

• an ONNX export of the trained model (best model), stored in weights/model.onnx.
• a Parquet file containing predictions on the full dataset, stored in predictions.parquet.
• a TensorRT engine export of the trained model, stored in weights/model.engine.
• metrics JSON files for each exported model format, stored in metrics_*.json:
  • metrics.json: metrics for the original PyTorch model
  • metrics_onnx.json: metrics for the ONNX exported model
  • metrics_tensorrt.json: metrics for the TensorRT exported model