Evaluations of training paradigms in neural image captioning

• Main Author: Lee, Si Min
• Other Authors: Zhang Hanwang
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2019
• Subjects: Engineering; Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/136508
• Institution: Nanyang Technological University

This project presents an implementation of 2 neural image captioning model which shall achieve results comparable to state of the art models. The 2 models are composed of different training paradigm, cross entropy training or self-critical training: • Model-1: Bottom-up attention with Self-Critical Training • Model-2: Bottom-up attention with cross entropy Training There will be a comparison and evaluation of the results against each model and related neural image captioning models to determine the best performing model.