Skip to content
ModelListedExternal
Download

Rfdetr Basketball Player Ball Referee Detection

This model is a fine-tuned version of Roboflow/rf-detr-nano on an labeled basketball object detection dataset from Roboflow. Roboflow Dataset - Same Dataset as above but exported to HF -

Sourcehuggingface.co/koppolusameer/rfdetr-basketball-player-ball-referee-detection/tree/46c33088c790670a7e81e21e753fa368b2d77a70
KindOtherBase modelRoboflow/rf-detr-nanoVersionRevision 46c3308

Model source

Kind
Other
File size
115 MB
Base model
Roboflow/rf-detr-nano
Version
vRevision 46c3308
Updated
Sep 22, 2026
Downloads
50
Source
Hugging Face

Source description

rfdetr-basketball-player-ball-referee-detection

This model is a fine-tuned version of Roboflow/rf-detr-nano on an labeled basketball object detection dataset from Roboflow.

It achieves the following results on the evaluation set:

  • Loss: 11.6183
  • Map: 0.4681
  • Map 50: 0.7416
  • Map 75: 0.5072
  • Map Small: 0.3566
  • Map Medium: 0.6316
  • Map Large: -1.0
  • Mar 1: 0.1966
  • Mar 10: 0.5356
  • Mar 100: 0.5675
  • Mar Small: 0.4527
  • Mar Medium: 0.7196
  • Mar Large: -1.0
  • Map Ball: 0.1805
  • Map Player: 0.6228
  • Map Referee: 0.601

Model description

The model is the RF-DETR Nano Model (https://huggingface.co/Roboflow/rf-detr-seg-nano).

Intended uses & limitations

Model is finetuned to detect players, ball, and referee in basketball broadcast images. Below are a few limitations

  • The ball is only a few pixels in size and often gets mixed with the background and confuses the model (as described by mAP results)
  • Images are mostly from Celtics games and hence there may be a lack of training data diversity

Training and evaluation data

Training procedure

  • Model was trained for 10 epochs and with the same train, valid, and test split as in the dataset
  • Training data was randomly augmented before finalizing the train set, where the augmentations done on an image could include at least one of
    • Horizontal flipping
    • Brightness and Contrast changes
    • Saturation changes
    • Blurring
    • Shifting, scaling, and Rotation

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: cosine
  • num_epochs: 10

Training results

Training LossEpochStepValidation LossMapMap 50Map 75Map SmallMap MediumMap LargeMar 1Mar 10Mar 100Mar SmallMar MediumMar LargeMap BallMap PlayerMap Referee
4.91431.023211.31880.2210.39770.23260.18070.317-1.00.09150.34390.45610.32190.6726-1.00.01660.39830.2481
4.18712.046411.53110.34670.58330.37420.22840.5233-1.00.15250.45230.51930.4020.6907-1.00.04320.54130.4554
3.70043.069611.22140.40640.65340.46680.28840.5955-1.00.16530.49260.5350.41490.7117-1.00.07350.58260.563
3.42664.092811.36120.45110.73270.4880.33650.6067-1.00.18860.50890.54510.4330.7104-1.00.17930.59470.5794
3.30025.0116011.26770.45970.74570.49720.34270.6219-1.00.18660.51720.55190.4340.7134-1.00.18260.60460.5919
2.99166.0139211.55050.44920.70710.49810.33620.6311-1.00.19110.54060.57010.45280.7196-1.00.12640.61570.6054
2.78237.0162411.57640.46810.73370.50750.35460.633-1.00.20040.53240.56560.4540.7156-1.00.18620.61790.6001
2.56298.0185611.54850.46560.74180.49260.35170.6307-1.00.19590.5290.56540.4580.7161-1.00.18160.6230.5922
2.35309.0208811.61030.47060.74540.52140.35970.6312-1.00.19880.53650.56870.45440.7186-1.00.18860.62060.6025
2.756110.0232011.61820.46740.74150.50580.35550.6313-1.00.19710.53560.56760.45270.7197-1.00.17940.62130.6015

Framework versions

  • Transformers 5.10.2
  • Pytorch 2.11.0+cu128
  • Datasets 4.0.0
  • Tokenizers 0.22.2

Sources

1 source
Hugging FacePrimaryVersion 46c33088c790670a7e81e21e753fa368b2d77a70
huggingface.co/koppolusameer/rfdetr-basketball-player-ball-referee-detection/tree/46c33088c790670a7e81e21e753fa368b2d77a70

Verified Sep 9

Model artifacts

1 artifact

model.safetensors

safetensors · 115 MB · SHA-256 c141ac1f05ff…6c35 · Hugging Face

Download

Source excerpts

2 excerpts
ReadmeSep 9huggingface.co/koppolusameer/rfdetr-basketball-player-ball-referee-detection

--- library_name: transformers license: apache-2.0 base_model: Roboflow/rf-detr-nano tags: - generated_from_trainer model-index: - name: rfdetr-basketball-player-ball-referee-detection results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # rfdetr-basketball-player-ball-referee-detection This model is a fine-tuned version of [Roboflow/rf-detr-nano](https://huggingface.co/Roboflow/rf-detr-nano) on an labeled basketball object detection dataset from Roboflow. * Roboflow Dataset - https://universe.roboflow.com/roboflow-jvuqo/basketball-player-detection-3-ycjdo/dataset/18 * Same Dataset as above but exported to HF - https://huggingface.co/datasets/koppolusameer/basketball-coco-20260416 It achieves the following results on the evaluation set: - Loss: 11.6183 - Map: 0.4681 - Map 50: 0.7416 - Map 75: 0.5072 - Map Small: 0.3566 - Map Medium: 0.6316 - Map Large: -1.0 - Mar 1: 0.1966 - Mar 10: 0.5356 - Mar 100: 0.5675 - Mar Small: 0.4527 - Mar Medium: 0.7196 - Mar Large: -1.0 - Map Ball: 0.1805 - Map Player: 0.6228 - Map Referee: 0.601 ## Model description The model is the RF-DETR Nano Model (https://huggingface.co/Roboflow/rf-detr-seg-nano). ## Intended uses & limitations Model is finetuned to detect players, ball, and referee in basketball broadcast images. Below are a few limitations * The ball is only a few pixels in size and often gets mixed with the background and confuses the model (as described by mAP results) * Images are mostly from Celtics games and hence there may be a lack of training data diversity ## Training and evaluation data * Train, Validation, and Test Splits can be found here - https://huggingface.co/datasets/koppolusameer/basketball-coco-20260416 ## Training procedure * Model was trained for 10 epochs and with the same train, valid, and test split as in the dataset * Training data was randomly augmented before finalizing the train set, where the augmentations done on an image could include at least one of * Horizontal flipping * Brightness and Contrast changes * Saturation changes * Blurring * Shifting, scaling, and Rotation ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 2e-05 - train_batch_size: 2 - eval_batch_size: 2...

NoteSep 9huggingface.co/koppolusameer/rfdetr-basketball-player-ball-referee-detection

Source context: 50 downloads · 0 likes · Pipeline object-detection · Library transformers · Repo koppolusameer/rfdetr-basketball-player-ball-referee-detection