/resolve/main/ ) : Figure . Image of < . Model Card for BioCLIP 2 BioCLIP 2 is a foundation model for biology organismal images. It is trained on TreeOfLife 200M on the basis of a CLIP model (ViT 14/L) pre trained on LAION 2B. BioCLIP 2 yields state of the art performance in recognizing various species. More importantly, it demonstrates emergent properties beyond species classification after extensive hierarchical contrastive training. Model Details Model Description Foundation models trained at scale exhibit emergent properties beyond their initial training objectives. BioCLIP 2 demonstrates such emergence beyond species classification by scaling up the hierarchical contrastive training proposed by BioCLIP. The model is trained on TreeOfLife 200M (the largest and most diverse available dataset of biology images). We evaluate BioCLIP 2 on a diverse set of biological tasks. Through training at scale, BioCLIP 2 improves species classification by 18.1% over BioCLIP. More importantly, we demonstrate that BioCLIP 2 generalizes to diverse biological questions beyond species classification solely through species level supersvision. Further analysis reveals that BioCLIP 2 acquires two emer…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy