|
--- |
|
language: |
|
- en |
|
metrics: |
|
- accuracy |
|
pipeline_tag: image-classification |
|
tags: |
|
- computer vision |
|
- flower |
|
- image classification |
|
- resnet50 |
|
--- |
|
|
|
|
|
# flower_image_classification_ResNet50_v1.0 |
|
|
|
This model is a fine-tuned version of Keras ResNet50 on the tf_flower dataset (https://www.tensorflow.org/datasets/catalog/tf_flowers). |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.7941 |
|
- Accuracy: 0.8571 |
|
|
|
## Model description |
|
|
|
A slightly customized image classification model for classify 5 labels of flowers ('daisy', 'dandelion', 'roses', 'sunflowers', 'tulips') |
|
|
|
## Intended uses & limitations |
|
|
|
This model is fined tune solely for flower image classification. |
|
|
|
## Training and evaluation data |
|
|
|
Training and testing data is splitted into 80:20 portion. |
|
Total data : 3670 files belonging to 5 classes |
|
Training data : 2753 files (80%) |
|
Validation data : 917 files (20%) |
|
|
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 1e-03 |
|
- train_batch_size: 8 |
|
- eval_batch_size: 8 |
|
- seed: 1 |
|
- optimizer: Adam |
|
- loss: categorical_crossentropy |
|
- num_epochs: 5 |
|
|
|
### Fine-Tuning Results |
|
|
|
| Epoch | Step | Training Loss | Training Accuracy | Validation Loss | Validation Accuracy| |
|
|:-----:|:-----:|:---------------:|:-----------------:|:---------------:|:------------------:| |
|
| 1.0 | 345 | 13.9143 | 0.6478 | 0.5310 | 0.8288 | |
|
| 2.0 | 690 | 0.2639 | 0.9161 | 0.6046 | 0.8419 | |
|
| 3.0 | 1035 | 0.1369 | 0.9539 | 0.5483 | 0.8561 | |
|
| 4.0 | 1380 | 0.0863 | 0.9703 | 0.5699 | 0.8659 | |
|
| 5.0 | 1725 | 0.0686 | 0.9837 | 0.7941 | 0.8571 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.41.2 |
|
- Pytorch 2.3.0 |
|
- opencv-contrib-python-4.10.0.82 |
