Image analytics using artificial intelligence : pose guided human image generation with deep neural network
Given the widespread problems of gelatinization and texture loss in the current image generation, a pose-guided human image generation model with RFB (Receptive Field Block) and SE (Squeeze-and-Excitation) Module added is proposed. This model uses GAN (Generative Adversarial Network) for training. I...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2021
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/150150 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Given the widespread problems of gelatinization and texture loss in the current image generation, a pose-guided human image generation model with RFB (Receptive Field Block) and SE (Squeeze-and-Excitation) Module added is proposed. This model uses GAN (Generative Adversarial Network) for training. It is used in pose integration and image refinement. Inspired by the attention mechanism of the channel feature, it is advisable to put the SE Module into the encoder block of Generator1 and Generator2. In addition, it also puts RFB into the forward layer of Generator1 and Generator2. It aims at enhancing the robustness of the image generation model and improve the quality of the produced image. The experimental results illustrate that the proposed model can obtain a higher evaluation score. It generates a more realistic and delicate human pose image that conforms to visual perception. |
|---|