# Question on text input in image captioning

**URL:** <https://discuss.huggingface.co/t/question-on-text-input-in-image-captioning/27198>\
**Category:** Beginners\
**Created:** [December 4, 2022, 12:20am UTC](https://discuss.huggingface.co/t/question-on-text-input-in-image-captioning/27198 "2022-12-04T00:20:09Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![shantanuacharya](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/shantanuacharya/32/63659_2.png) [@shantanuacharya](https://discuss.huggingface.co/u/shantanuacharya)\
**Post date:** [December 4, 2022, 12:20am UTC](https://discuss.huggingface.co/t/question-on-text-input-in-image-captioning/27198/1 "2022-12-04T00:20:09Z")

</div>

I was going through this [blog](https://huggingface.co/nlpconnect/vit-gpt2-image-captioning) on image captioning.

According to the blog, the `VisionEncoderDecoderModel` uses this kind of architecture (shown below) where the image embeddings are fed to the decoder via the _Encoder-Decoder Self-Attention_.  
 ![Screenshot 2022-12-03 at 7.17.36 PM](https://us1.discourse-cdn.com/hellohellohello/original/2X/f/f17f4ece49bddeebaab3d84681b2df4bccf427f9.png)

What I wanted to ask is that if the input is only image then what are we feeding as input (denoted by _Text Caption_ in the image) to the Transformer decoder?