Merge pull request #394 from the-silent-geek/Heat_sink_defect_detection

Heat sink defect detection
abhisheks008 · Dec 24, 2023 · ff75313 · ff75313
2 parents ccc4849 + 1280916
commit ff75313
Show file tree

Hide file tree

Showing 10 changed files with 1,813 additions and 0 deletions.
diff --git a/Heat sink defect detection/dataset/README.md b/Heat sink defect detection/dataset/README.md
@@ -0,0 +1 @@
+### Dataset Link : https://www.kaggle.com/datasets/kaifengyang/heat-sink-surface-defect-dataset
diff --git a/Heat sink defect detection/images/01.png b/Heat sink defect detection/images/01.png
diff --git a/Heat sink defect detection/images/02.png b/Heat sink defect detection/images/02.png
diff --git a/Heat sink defect detection/images/03.png b/Heat sink defect detection/images/03.png
diff --git a/Heat sink defect detection/images/04.png b/Heat sink defect detection/images/04.png
diff --git a/Heat sink defect detection/images/05.png b/Heat sink defect detection/images/05.png
diff --git a/Heat sink defect detection/images/06.png b/Heat sink defect detection/images/06.png
diff --git a/Heat sink defect detection/model/README.md b/Heat sink defect detection/model/README.md
@@ -0,0 +1,26 @@
+**HEAT SINK DEFECT DETECTION**
+
+**GOAL**
+
+To detect stains and scratches on the given heat sink images.
+
+**DATASET**
+
+https://www.kaggle.com/datasets/kaifengyang/heat-sink-surface-defect-dataset
+
+**DESCRIPTION**
+
+This project aims to develop a robust defect detection system utilizing U-Net architecture models, an effective neural network design for image segmentation tasks. By harnessing deep learning techniques, this project strives to accurately identify and localize defects within heat sinks.
+
+
+**MODELS USED**
+
+U-net - U-Net, a deep learning architecture, excels in image segmentation tasks. Its unique design incorporates encoder-decoder pathways, ideal for precise localization, widely applied in medical imaging and object detection.
+
+Resnet50 - ResNet-50 is a 50-layer convolutional neural network (48 convolutional layers, one MaxPool layer, and one average pool layer). Residual neural networks are a type of artificial neural network (ANN) that forms networks by stacking residual blocks. It excels in image recognition tasks, offering high accuracy and efficiency in deep learning models.
+
+Vgg16 - VGG16 is a deep convolutional neural network renowned for its 16 layers, characterized by a simple yet effective architecture. Its design, with small receptive fields and stacked layers, excels in image classification tasks, making it a popular choice for feature extraction and transfer learning in computer vision applications.
+
+**ACCURACIES**
+
+All 3 models gave accuracies of 97.8% on training upon 40 epochs of batch size 32.
diff --git a/Heat sink defect detection/model/model.ipynb b/Heat sink defect detection/model/model.ipynb
diff --git a/Heat sink defect detection/requirements.txt b/Heat sink defect detection/requirements.txt
@@ -0,0 +1,9 @@
+imageio==2.26.0
+keras==2.15.0
+matplotlib==3.7.2
+numpy==1.24.3
+pandas==2.0.3
+scikit-learn==1.3.0
+seaborn==0.12.2
+segmentation-models==1.0.1
+tensorflow==2.15.0
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		### Dataset Link : https://www.kaggle.com/datasets/kaifengyang/heat-sink-surface-defect-dataset