Merge pull request #72 from pavitraag/main

Added Human Detection Project

Neural Networks/Human Detection Project/Dataset/Readme.md

@@ -0,0 +1,7 @@
+The link for the dataset used in this project: https://www.kaggle.com/datasets/constantinwerner/human-detection-dataset
+
+- Dataset contains CCTV footage images(as indoor as outdoor), a half of them w humans and a half of them is w/o humans. Images is marked as follow:
+
+0_n.png or 1_n.png
+
+- The first digit is a class of image, 0 means a scene without humans, and 1 means a scene with humans.

Neural Networks/Human Detection Project/Readme.md

@@ -0,0 +1,74 @@
+# Human Detection Project
+
+## PROJECT TITLE
+
+Human Detection Project
+
+## GOAL
+
+To create a DL model which will identify the humans in given image.
+
+## DATASET
+
+The dataset used for this project can be found at [link to dataset](https://www.kaggle.com/datasets/constantinwerner/human-detection-dataset). 
+
+## DESCRIPTION
+
+This project aims to identify the humans in the image. It is trained on the dataset containing CCTV footage images(as indoor as outdoor).
+
+## WHAT I HAD DONE
+
+1. Data collection: From the link of the dataset given above. 
+2. Data preprocessing: Created a Image generators which helped to generate more images in order to increase the accuracy.
+3. Model selection: Chose traditional CNN along with Image detection architecture VGG16, Inception and Xception for Image detection.
+4. Comparative analysis: Compared the accuracy score of all the models.
+
+## MODELS USED
+
+1. CNN
+2. VGG16
+3. Inception
+4. Xception
+
+
+## LIBRARIES NEEDED
+
+The following libraries are required to run this project:
+
+- numpy==1.24.3
+- pandas==1.5.0
+- matplotlib==3.6.0
+- tensorflow==2.6.0
+
+
+## EVALUATION METRICS
+
+The evaluation metrics I used to assess the models:
+
+- Accuracy 
+- Loss
+- Confusion Matrix
+
+## RESULTS
+Results on Val dataset:
+
+| Model      | Accuracy | Loss    |
+|------------|----------|---------|
+| CNN    | 0.756     | 0.591   |
+| VGG16    | 0.859     | 0.304    |
+| Inception    | 0.886     | 0.33    |
+| Xception    | 0.843    | 0.481    |
+| Fine-tuned Inception    | 0.902    | 0.312    |
+
+
+## CONCLUSION
+Based on results we can draw following conclusions:
+1. CNN Model: The CNN model achieved an accuracy of 75.6% and a loss of 0.591. While it demonstrates reasonable performance, there is room for improvement compared to the other models.
+
+2. VGG16 Model: The VGG16 model outperformed the CNN model with an accuracy of 85.9% and a lower loss of 0.304. It shows the effectiveness of using a pre-trained model like VGG16 for human detection, achieving a higher accuracy and lower loss than the CNN model.
+
+3. Inception Model: The Inception model achieved an accuracy of 88.6% and a loss of 0.33. It performs well, with a higher accuracy than both the CNN and VGG16 models. This suggests that the Inception architecture is effective in capturing human features and distinguishing them from non-human objects.
+
+4. Xception Model: The Xception model achieved an accuracy of 84.3% and a loss of 0.481. While it performs decently, it is slightly behind the VGG16 and Inception models in terms of accuracy.
+
+5. Fine-tuned Inception: The fine-tuned Inception model showed further improvement with an accuracy of 90.2% and a loss of 0.312. Fine-tuning the Inception model likely helped to adapt it more specifically to the task of human image detection, resulting in increased accuracy.

Neural Networks/Human Detection Project/Requirements.txt

@@ -0,0 +1,5 @@
+numpy==1.24.3
+pandas==1.5.0
+matplotlib==3.6.0
+tensorflow==2.6.0
+seabor==0.11.2