@nielsr Hi, I’m currently trying to build an OCR (text-recognition model), which gets the visual and contextual features via a transformer architecture. Currently my architecture looks like this: Visual features: DeiT FeatureExtractor (can it be that this is just a preprocessing instance for the …

Extract visual and contextual features from images

Felix92 August 27, 2021, 1:23pm 6

Input Example after FeatureExtractor:

Topic		Replies	Views
Using trasnsformer to get image features 🤗Transformers	3	3393	March 20, 2024
Why TrOCR processor has a feature extractor? Beginners	8	1476	November 25, 2021
Using Huggingface for computer vision (Tensorflow)? 🤗Transformers	3	446	June 2, 2025
Img2seq model with pretrained weights Beginners	7	1264	November 18, 2021
Image Features as Model Input Beginners	2	964	November 18, 2020