--- language: en license: mit tags: - vision - image-to-text - image-captioning model_name: microsoft/git-base pipeline_tag: image-to-text library_name: transformers datasets: - ayoubkirouane/One-Piece-anime-captions --- # Model Details + **Model Name**: Git-base-One-Piece + **Base Model**: Microsoft's "git-base" model + **Model Type**: Generative Image-to-Text (GIT) + **Fine-Tuned** On: 'One-Piece-anime-captions' dataset + **Fine-Tuning Purpose**: To generate text captions for images related to the anime series "One Piece." ## Model Description **Git-base-One-Piece** is a fine-tuned variant of Microsoft's **git-base** model, specifically trained for the task of generating descriptive text captions for images from the **One-Piece-anime-captions** dataset. The dataset consists of **856 {image: caption}** pairs, providing a substantial and diverse training corpus for the model. The model is conditioned on both CLIP image tokens and text tokens and employs a **teacher forcing** training approach. It predicts the next text token while considering the context provided by the image and previous text tokens. ![image/jpeg](https://cdn-uploads.huggingface.co/production/uploads/6338c06c107c4835a05699f9/N_yNK2tLabtwmSYAqpTEp.jpeg) ## Limitations + The quality of generated captions may vary depending on the complexity and diversity of images from the **One-Piece-anime-captions** dataset. + The model's output is based on the data it was fine-tuned on, so it may not generalize well to images outside the dataset's domain. Generating highly detailed or contextually accurate captions may still be a challenge. ## Usage ```python # Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-to-text", model="ayoubkirouane/git-base-One-Piece") ``` **or** ```python # Load model directly from transformers import AutoProcessor, AutoModelForCausalLM processor = AutoProcessor.from_pretrained("ayoubkirouane/git-base-One-Piece") model = AutoModelForCausalLM.from_pretrained("ayoubkirouane/git-base-One-Piece") ```