ElstnerAnalytics/chemical-documents

--- license: cc-by-4.0 task_categories: - image-segmentation - image-feature-extraction language: - en tags: - chemistry size_categories: - 1K<n<10K --- The dataset consists of CC-BY-4.0 licensed open access papers and the annotations on-top are provided under the same license. Train, test and validation splits are provided as separate folders containing the image files and a single json file containing the annotations in COCO format. The classes are close to the [TFT-ID](https://huggingface.co/datasets/yifeihu/TFT-ID-1.0-coco) dataset and extend into 3 chemistry specific labels. `chem_reaction` contains everything with reaction arrows, `chem_structures` are structural formulas without reactions and `chem_table` is used for tables containing structural formula. The dataset is also available on [Roboflow](https://app.roboflow.com/elstner-analytics/ttfc/2), allowing to directly convert the annotatios into different formats.