Unified Mentor Private Limited
## Table of Contant
-
About this Project.
-
What DataSet cointains?
-
Main library to be used.
-
Visualizations/Chart.
-
Conclusion.
-
Acknowledgment.
Build a system that can identify the animal in a given image. Explore the data set and identify an appropriate solution.
- Dataset exploration
- Image preprocessing
- Transfer Learning with MobileNetV2
- Model evaluation
- Prediction on new images
dataset/ │ ├── Bear/*.jpg ├── Bird/*.jpg ├── Cat/*.jpg ├── Cow/*.jpg ├── Deer/*.jpg ├── Dog/*.jpg ├── Dolphin/*.jpg ├── Elephant/*.jpg ├── Giraffe/*.jpg ├── Horse/*.jpg ├── Kangaroo/*.jpg ├── Lion/*.jpg ├── Panda/*.jpg ├── Tiger/*.jpg └── Zebra/*.jpg Main Libraries Used
sys – for understanding venv and version.
json – for saving report of history of models evaluation.
NumPy – For efficient numerical computations, array operations, and mathematical functions.
Matplotlib & Seaborn – For creating visualizations such as histograms, heatmaps, and correlation plots to explore data patterns and relationships.
scikit-learn (sklearn) – The main machine learning library used for:
- 🎯 Model evaluation (
confusion_matrix,Classification_report)
- 🎯 importing images (
image_dataset_from_directory) - 🎯 Creating model (
layers,keras) - 🎯 Loading saved model (
load_model)
Insights & Visualizations/Chart.
precision recall f1-score support Bear 0.91 0.87 0.89 23 Bird 1.00 0.90 0.95 21 Cat 0.89 0.96 0.92 25 Cow 0.87 0.92 0.89 36 Deer 1.00 0.82 0.90 34 Dog 1.00 0.78 0.88 18 Dolphin 0.92 0.96 0.94 24 Elephant 1.00 0.84 0.91 25 Giraffe 1.00 0.93 0.96 27 Horse 0.79 0.96 0.87 27 Kangaroo 0.76 0.92 0.83 24 Lion 0.97 0.90 0.93 31 Panda 0.85 0.96 0.90 23 Tiger 0.95 1.00 0.98 21 Zebra 0.97 1.00 0.98 29 accuracy 0.91 388 macro avg 0.92 0.91 0.92 388 weighted avg 0.92 0.91 0.92 388 Model: "sequential_1" --------------------------------------------------------------- Layer (type) | Output Shape | Param # --------------------------------------------------------------- rescaling_1 (Rescaling) | (None, 224, 224, 3)| 0 mobilenetv2_1.00_224 (Functional) | (None, 7, 7, 1280) | 2,257,984 global_average_pooling2d_1 | (None, 1280) | 0 dense_2 (Dense) | (None, 120) | 153,720 dropout_1 (Dropout) | (None, 120) | 0 dense_3 (Dense) | (None, 15) | 1,815 --------------------------------------------------------------- Total params: 2,724,591 Trainable params: 155,535 Non-trainable params: 2,257,984
- Training Accuracy: 1.0%
- Validation Accuracy: 91.4%
This project is dedicated to applying machine learning techniques to understand and detect Animal Classification. Sincere thanks to Unified Mentor Private Limited for providing the opportunity and platform to carry out this work. Appreciation is also extended to the open-source community for developing the powerful tools and libraries that made this project possible.
Created with 🧠 by Kushang Shah