Wildlife Sentinel Wildlife Sentinel is the public Hugging Face dataset package for the Savanna Sentinel project: a multimodal, tool grounded biodiversity monitoring dataset built from Snapshot Serengeti camera trap events and public environmental layers. This Hub package intentionally contains the train, validation, and evaluation components only, plus supporting artifacts needed to understand and reproduce those splits. Demo replay inputs, hidden demo labels, raw source caches, full environmental raw downloads, and product local artifacts remain local in the Savanna Sentinel workspace. The package contains train/validation/evaluation camera trap image bursts, training JSONL files, filtered manifests/features, schemas, scripts, and audit reports. It is designed for fine tuning and evaluating Gemma style vision language models that classify camera trap events, route uncertain cases for review, and answer ecological questions through tool calling over grounded data tables. What Is Included Image And Event Data data/mirrored images/snapshot serengeti 50gb/train core/ data/mirrored images/snapshot serengeti 50gb/train blank/ data/mirrored images/snapshot serengeti 50gb/val event/ data/…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy