Skip to content

meetrajsinhJ/IKTD_IKLEUS_FeedbackSystem

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

32 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

IKTD_IKLEUS_FeedbackSystem

Bewertung von Technische Zeichnungen

Table of Contents

  1. Dataset
  2. Data Augmentation
  3. Clustering for Exploratory Data Analysis
  4. Computer Vision Operations
  5. Image Classification
  6. Image Segmentation
  7. Feedbacksystem Module

1.DATASET

4 different types of Technical Drawings along with their respective Evaluation Table:- a) Schraube b) Baugruppen c) Toleranz d) Gussaufgabe

Screenshot 2024-12-10 at 15 05 52

2.DATA AUGMENTATION

a). Increasing the the dataset by implementating various data Augmentation Techniques: Scale variation , Brightness and Contrast , Rotation etc.

Screenshot 2024-12-10 at 15 15 05

b). Adding noise and reconstruction using Variational AutoEncoders

reconstructed vaeimages

3.CLUSTERING

Appling k-means Clustering algorithm on the dataset.

a) Clustering on Corrected and Not-Corrected Images

Screenshot 2024-12-10 at 15 21 03

b) Clustering on Entire Dataset

Screenshot 2024-12-10 at 15 21 18

4.COMPUTER VISION OPERATIONS:

a). Canny Edge Detection:

Screenshot 2024-12-10 at 15 23 31 Screenshot 2024-12-10 at 15 23 40

b). SIFT feature Matching and Homography Matrix Calculations:

Screenshot 2024-12-10 at 15 24 49

5.IMAGE CLASSIFICATION:

-> Classification of the images based on the drawing type:

Finetuning ResNet50 , EfficicentNet Models on this custom Dataset. Models along with the Training Scripts are uploaded in Image Classification directory.

6.IMAGE SEGMENTATION:

-> Segementing the drawing and the text box on the Image.

Trained a yolov8 model on custom dataset.

result

7.FEEDBACK SYSTEM MODULE:

a). Overview of the system Architecture

Screenshot 2024-11-17 at 18 01 01

b). Dataset Preparation:

Screenshot 2024-12-10 at 15 36 30

c). Model Fusion:

-Combining Visual Features with Text Analysis along with Domain Knowledge to generate feedback on the drawing. -Training the fusion Model on the custom annotated Dataset for 1000 Epochs.

training_loss training_metrics

d). Output:

-> Deployment using FASTAPI on local host for inference check

inference_visualization inference_visualization9 inference_visualization10

Contact Author: [email protected]

Institut für Konstruktionstechnik und Technisches Design https://www.iktd.uni-stuttgart.de

About

Bewertung von Technische Zeichnungen

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published