Skip to content

Latest commit

 

History

History
111 lines (104 loc) · 5.09 KB

README.md

File metadata and controls

111 lines (104 loc) · 5.09 KB

One Piece Image Recognition

A machine learning model that can identify images of One Piece characters

Requirements:

Troubleshooting:

Setup and Usage:

  1. Clone and enter the repo
git clone git@github.com:greshbasic/op-image-recognition.git
cd op-image-recognition
  1. Create a venv and install requirements
python3 -m venv <name>
source <name>/bin/activate
pip install -r requirements.txt
  1. Run the main.py file
python3 main.py

What to Expect:

  • A lot of information is logged to the console, mostly warnings. You can ignore those as they aren't going to negatively impact the model. However, if you want to build upon my work it may eventually become something you need to look into.
  • On my mid-range PC build it takes about 2 minutes for the model to fully train.
  • The model averages between 70-75% validation accuracy and is slightly overfit. 
    • Epoch 1/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 26s 58ms/step - accuracy: 0.0551 - loss: 2.9549 - val_accuracy: 0.0739 - val_loss: 2.8883 - learning_rate: 0.0010
Epoch 2/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 8s 26ms/step - accuracy: 0.0798 - loss: 2.8492 - val_accuracy: 0.1483 - val_loss: 2.6521 - learning_rate: 0.0010
Epoch 3/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 23ms/step - accuracy: 0.1881 - loss: 2.5654 - val_accuracy: 0.2438 - val_loss: 2.3851 - learning_rate: 0.0010
Epoch 4/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 23ms/step - accuracy: 0.2864 - loss: 2.2582 - val_accuracy: 0.3511 - val_loss: 2.1002 - learning_rate: 0.0010
Epoch 5/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 23ms/step - accuracy: 0.3707 - loss: 1.9949 - val_accuracy: 0.4432 - val_loss: 1.8730 - learning_rate: 0.0010
Epoch 6/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 23ms/step - accuracy: 0.4450 - loss: 1.7575 - val_accuracy: 0.4818 - val_loss: 1.7609 - learning_rate: 0.0010
Epoch 7/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 22ms/step - accuracy: 0.5161 - loss: 1.5475 - val_accuracy: 0.5028 - val_loss: 1.6314 - learning_rate: 0.0010
Epoch 8/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 22ms/step - accuracy: 0.5589 - loss: 1.3771 - val_accuracy: 0.5665 - val_loss: 1.4851 - learning_rate: 0.0010
Epoch 9/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 22ms/step - accuracy: 0.6136 - loss: 1.2008 - val_accuracy: 0.5761 - val_loss: 1.4158 - learning_rate: 0.0010
Epoch 10/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 22ms/step - accuracy: 0.6428 - loss: 1.1050 - val_accuracy: 0.6074 - val_loss: 1.3171 - learning_rate: 0.0010
Epoch 11/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 22ms/step - accuracy: 0.6845 - loss: 0.9817 - val_accuracy: 0.6313 - val_loss: 1.2414 - learning_rate: 0.0010
Epoch 12/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 22ms/step - accuracy: 0.7127 - loss: 0.8847 - val_accuracy: 0.6097 - val_loss: 1.2567 - learning_rate: 0.0010
Epoch 13/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 22ms/step - accuracy: 0.7490 - loss: 0.7974 - val_accuracy: 0.6585 - val_loss: 1.1779 - learning_rate: 0.0010
Epoch 14/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 23ms/step - accuracy: 0.7575 - loss: 0.7488 - val_accuracy: 0.6562 - val_loss: 1.1683 - learning_rate: 0.0010
Epoch 15/15
312/312 ━━━━━━━━━━━━━━━━━━━━ 7s 22ms/step - accuracy: 0.7769 - loss: 0.6842 - val_accuracy: 0.6784 - val_loss: 1.1223 - learning_rate: 0.0010
  • Users will be shown 9 random images and the model's classification of each, alongside the training history
Special Thanks: I want to give a huge thanks to Ibrahim Serouis on Kaggle for creating the One Piece character image data set!