datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
rf100-vl-results
RF100-VL campaign artifacts
Raw artifacts from RF100-VL benchmark campaigns run with LibreYOLO: the
per-dataset training configs, per-epoch metrics, logs, GPU telemetry, scoring
inputs and submissions. Published so a result can be checked rather than
believed.
Protocol: fine-tune one checkpoint per dataset across the 100 RF100-VL
datasets, score each on its test split with pycocotools at maxDets 500, and
report the unweighted mean AP50:95. Epochs, batch size, seed and selection… See the full description on the dataset page: https://huggingface.co/datasets/LibreYOLO/rf100-vl-results.rf100-vl
RF100-VL, packaged for one download
This is a temporary redistribution of Roboflow's RF100-VL benchmark, published so that benchmark runs can fetch the data in one step instead of a hundred. It is not an original work, and if Roboflow asks us to take it down we will, immediately and without argument.
Everything here comes from Roboflow 100-VL. The benchmark, the datasets, the splits and the cleaning logic are theirs. We changed none of it.
Why this exists
RF100-VL… See the full description on the dataset page: https://huggingface.co/datasets/LibreYOLO/rf100-vl.rf100-vlrf100-vl-datasetsrf100-vlrf100-vl-fsodrf100-vl-backuprf100-vl