Simple RNN training for classification task of 3 signal: Sine, Square, Triangle.
Simple RNN training for sine wave estimation.
Comparison of RNN model and GRU model for classification task of 3 signal: Sine, Square and Triangle, after 100 epoch training.
| Model | Accuracy |
|---|---|
| RNN Model | 0.9315 |
| GRU Model | 0.9383 |
Comparison of RNN model and GRU model for regression task of sine wave estimation after 100 epoch training.
| Model | loss |
|---|---|
| RNN Model | 0.0027 |
| GRU Model | 0.0026 |
Generate data for ball move direction
GRU implementation from scratch + inference
LSTM implementation from scratch + inference
-
Generate data for ball move direction
-
Classification of direction using RNN, GRU and LSTM
- 09- Video_Classification_CRNN.ipynb(train)
- inference.py
- models.py (gru, lstm, rnn)
- load_video.py
- requirements.txt
Backbone: ResNet50V2 and my vgg base model for feature extraction
RNN modules: RNN, GRU and LSTM are tested
The performance of GRU module was better than other madules
Dataset contains videos from 2 classes
Due to insufficient data, the training was not done well. but this project can be used for other video classification tasks using CRNNs
- Video classificatio nusing CRNN on ucf101_top5 dataset
Backbone: my vgg base model for feature extraction
RNN modules: RNN, GRU are tested
The performance of GRU module was better than RNN madules
- ucf101_top5 dataset containing 573 video from 5 classes
| Model | Val Accuracy |
|---|---|
| RNN Model | 0.87 |
| GRU Model | 0.94 |