https://github.com/bhuvan-s-prasad/bird_classification-resnet

A deep learning project utilizing ResNet-50 to classify images of 100 different bird species. This model uses transfer learning with data augmentation, learning rate scheduling, early stopping, and cross-entropy loss to achieve accurate classification.
https://github.com/bhuvan-s-prasad/bird_classification-resnet

css flask html javascript machine-learning neural-network python python3 pytorch resnet-50 tensorflow

Last synced: 5 months ago
JSON representation

A deep learning project utilizing ResNet-50 to classify images of 100 different bird species. This model uses transfer learning with data augmentation, learning rate scheduling, early stopping, and cross-entropy loss to achieve accurate classification.

• Host: GitHub
• URL: https://github.com/bhuvan-s-prasad/bird_classification-resnet
• Owner: Bhuvan-S-prasad
• Created: 2024-11-11T06:20:23.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2025-03-22T11:44:49.000Z (over 1 year ago)
• Last Synced: 2025-03-31T02:21:32.970Z (over 1 year ago)
• Topics: css, flask, html, javascript, machine-learning, neural-network, python, python3, pytorch, resnet-50, tensorflow
• Language: Python
• Homepage: https://web-production-4c59.up.railway.app/
• Size: 172 MB
• Stars: 0
• Watchers: 1
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Bird Species Classification using convolutional neural network - resnet50

A deep learning project utilizing ResNet-50 to classify images of 100 different bird species. This model uses transfer learning with data augmentation, learning rate scheduling, early stopping, and cross-entropy loss to achieve accurate classification.

## Table of Contents

- [Overview](#overview)

- [Requirements](#requirements)

- [Dataset](#dataset)

- [Classes](#Classes)

- [Model Architecture](#model-architecture)

- [Training Pipeline](#training-pipeline)

- [Prediction](#prediction)

- [Usage](#usage)

- [Results](#results)

- [Acknowledgement](#Acknowledgement)

## Overview

This project implements a bird species classifier using a modified ResNet-50 architecture trained on a custom dataset with 100 bird species. Data augmentation is applied to improve model generalization, and training metrics are logged to track model performance.

## Requirements

To set up and run this project, install the following packages:

- Python 3.x

- PyTorch

- torchvision

- scikit-learn

- pandas

- numpy

- seaborn

- matplotlib

- tqdm

- psutil

- PIL (Pillow)

```bash

pip install torch torchvision scikit-learn pandas numpy seaborn matplotlib tqdm psutil pillow

```

## Dataset

The dataset consists of three folders: `train`, `val`, and `test`, located in the `data_dir` directory. Each folder contains images categorized by bird species. This project uses a `[70-20-10]` split for training, validation, and testing.

## Classes

The dataset contains the following bird species along with an "unknown" class:

1. Black-footed Albatross  

2. Laysan Albatross  

3. Sooty Albatross  

4. Groove-billed Ani  

5. Crested Auklet  

6. Least Auklet  

7. Parakeet Auklet  

8. Rhinoceros Auklet  

9. Brewer Blackbird  

10. Red-winged Blackbird  

11. Rusty Blackbird  

12. Yellow-headed Blackbird  

13. Bobolink  

14. Indigo Bunting  

15. Lazuli Bunting  

16. Painted Bunting  

17. Cardinal  

18. Spotted Catbird  

19. Gray Catbird  

20. Yellow-breasted Chat  

21. Eastern Towhee  

22. Chuck-will's-widow  

23. Brandt's Cormorant  

24. Red-faced Cormorant  

25. Pelagic Cormorant  

26. Bronzed Cowbird  

27. Shiny Cowbird  

28. Brown Creeper  

29. American Crow  

30. Fish Crow  

31. Black-billed Cuckoo  

32. Mangrove Cuckoo  

33. Yellow-billed Cuckoo  

34. Gray-crowned Rosy Finch  

35. Purple Finch  

36. Northern Flicker  

37. Acadian Flycatcher  

38. Great Crested Flycatcher  

39. Least Flycatcher  

40. Olive-sided Flycatcher  

41. Scissor-tailed Flycatcher  

42. Vermilion Flycatcher  

43. Yellow-bellied Flycatcher  

44. Frigatebird  

45. Northern Fulmar  

46. Gadwall  

47. American Goldfinch  

48. European Goldfinch  

49. Boat-tailed Grackle  

50. Eared Grebe  

51. Horned Grebe  

52. Pied-billed Grebe  

53. Western Grebe  

54. Blue Grosbeak  

55. Evening Grosbeak  

56. Pine Grosbeak  

57. Rose-breasted Grosbeak  

58. Pigeon Guillemot  

59. California Gull  

60. Glaucous-winged Gull  

61. Heermann's Gull  

62. Herring Gull  

63. Ivory Gull  

64. Ring-billed Gull  

65. Slaty-backed Gull  

66. Western Gull  

67. Anna's Hummingbird  

68. Ruby-throated Hummingbird  

69. Rufous Hummingbird  

70. Green Violetear  

71. Long-tailed Jaeger  

72. Pomarine Jaeger  

73. Blue Jay  

74. Florida Jay  

75. Green Jay  

76. Dark-eyed Junco  

77. Tropical Kingbird  

78. Gray Kingbird  

79. Belted Kingfisher  

80. Green Kingfisher  

81. Pied Kingfisher  

82. Ringed Kingfisher  

83. White-breasted Kingfisher  

84. Red-legged Kittiwake  

85. Horned Lark  

86. Pacific Loon  

87. Mallard  

88. Western Meadowlark  

89. Hooded Merganser  

90. Red-breasted Merganser  

91. Mockingbird  

92. Nighthawk  

93. Clark's Nutcracker  

94. White-breasted Nuthatch  

95. Baltimore Oriole  

96. Hooded Oriole  

97. Orchard Oriole  

98. Scott's Oriole  

99. Ovenbird  

100. Brown Pelican  

101. Unknown

## Model Architecture

This project uses a pre-trained ResNet-50 architecture modified to classify 100 bird species:

- **Transfer Learning**: Pre-trained weights on ImageNet are used to initialize the model.

- **Final Layer Adjustment**: The final fully connected layer is modified to match the number of bird classes.

- **Optimization**: The model is trained using the Adam optimizer with weight decay to prevent overfitting.

## Training Pipeline

- **Data Augmentation**: Random transformations are applied to the training dataset.

- **Training Loop**: The model is trained with `CrossEntropyLoss` and the Adam optimizer, using `ReduceLROnPlateau` to adjust the learning rate based on validation loss.

- **Early Stopping**: Training stops early if validation loss does not improve for a set patience period.

Training metrics are saved to `training_log_fix_final.csv` for analysis, and the best model is saved as `best_model_fix_final.pth`.

## Prediction

The prediction pipeline allows inference on new bird images using the trained ResNet-50 model:

- The model expects images to be resized and normalized to match the training distribution.

- The `predict_image()` function loads the image, applies transformations, and returns the predicted bird species.

## application

- ```app.py``` built using flask and deployed in railway.app

## Usage

- **Prediction**: Use `prediction.py` to classify images. Update `image_path` and `model_path` accordingly.

## Results

The final model achieves notable accuracy on the validation set. Check the saved logs for per-epoch metrics and performance analysis.

finally deployed the flask application with the railway.app 

## Acknowledgement

This project is done under the guidance of [Dr Agughasi Victor Ikechukwu](https://github.com/Victor-Ikechukwu)

## Future Enhancement

- **Ensemble Approach**: Perform ensembling using at least 5 pre-trained models and compare their performance vs training from scratch.

- **Explainability**: Explore explainable approaches via GradCAM, SHAP & LIME.