Research Progress and Results in Image Dataset Analysis

Research progress and results in image dataset analysis including experiment outcomes, discussion on model performance, dataset analysis, and model training. The study covers topics such as analysis of kiwi leaf trips and spots, model ensemble techniques, teacher-student learning, and the effectiveness of different models in image classification tasks.

• Research Analysis
• Dataset
• Model Ensemble
• Image Classification
• Model Training

cazbi Follow

Uploaded on Oct 07, 2024 | 0 Views

Download Presentation

Please find below an Image/Link to download the presentation.

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.

Presentation Transcript

1. IIP 2022. 09. 01

2. Contents Preview Part.1 1. Experiment Result 2. Discussion Experiment in progress Part.2 1. 2. DATASET MODEL

3. Part.1 Preview

4. Preview Experiment Result Description of research progress and results Result Problem Kiwi Leaf Trips Kiwi Leaf Spot Spot Trips Kiwi Leaf Normal 4 / 26

5. Preview Discussion Description of research progress and results Discussion Test Dataset Best model weights Kiwi Dataset Training Validation Web Application

6. Preview Description of research progress and results Discussion (ResNet, DenseNet, EfficientNet) of ensemble Discussion MEAL(Multi-model Ensemble via Adversarial Learning) MEAL Teacher-Student learning , teacher , selection module distillation student network selection module teacher ensemble , one-hot/hard label distillation ResNet-50 ImageNet-1K (224 224 single crop) TOP-1 80% 6 / 26 Shen, Zhiqiang, Zhankui He, and Xiangyang Xue. "Meal: Multi-model ensemble via adversarial learning." Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 33. No. 01. 2019.

7. Preview Description of research progress and results Discussion Discussion DeiT (Data-efficient image Transformers) EfficientNet ViT CNN Model Class token + patch tokens Cross Entropy Loss CNN inductive bias Global Optimum 7 / 26 Training data-efficient image transformers & distillation through attention (2021)

8. Part.2 Experiment in progress

9. Graduation Paper Idea Description of research progress and results DATASET Self-collected Dataset + Open source Dataset Training set : 15,996 1. Training set + Validation set + Test set 2. Random Shuffle Open Dataset : PlantVillage Validation set : 2,002 3. Training set : Validation set = 7:3 17,000 : 3,000 Validation of model s effectiveness Test set : 2,002 9 / 22 9 / 22

10. Preview Discussion Description of research progress and results Discussion Best model weights Test Dataset & Web Application Kiwi Dataset Training Validation Best model weights

11. Preview Description of research progress and results Discussion Discussion (ResNet, DenseNet, EfficientNet) of ensemble ViT 11 / 26

12. IIP .