Image Caption Generator Using LSTM
Chalcheema Sasidhar, Madan Lal Saini, Medarametla Charan, Avula Venkata Shivanand, Vijay Mohan Shrimal · 2024
Generating image captions is a difficult task which implies capturing the main scene of an image and consequently labelling it with a natural language description. The paper aims to provide a unique image captioning using the NLP based techniques. State-of-the-art deep learning models, CNNs in particular as well as RNNs, are the main tools to be used in the suggested system through the analysis of the image features and the out coming sequences respectively. Initially, the system is converting images to output high level features which is used as an input to an autoregressive language model such as transformers to produces captions. The training process gets handled by the model that optimizes parameters to reach on expected image captions probability value. Captions are also produced by attention mechanism to bring the attention on each word of the caption. The performance assessment of the model is evaluated by BLEU (Bilingual Evaluation Understudy) and METEOR (Metric for Evaluation of Translation with Explicit Ordering) scores and these parameters are used for two aspects, quantitative and qualitative. These scores are utilized to plot the accuracy of generated captions against ground truth annotations.