Skip to content

Latest commit

 

History

History
45 lines (32 loc) · 2.38 KB

README.md

File metadata and controls

45 lines (32 loc) · 2.38 KB
Raw

Opti-Car

Car price predictor using a convolutional neural network. This was for the final assignment for the Neural Networks course.

By s3301419, s3814254, s3593673, s4170237

Dataset

The car pictures used for this project are from Nicolas Gervais' Car Connection Picture Dataset. This dataset originally consists of roughly 60,000 images which are all tagged with the suggested retail price (MSRP) of the cars.

Requirements

library version used scripts
python 3.7.3 *
tensorflow 2.2.0 opticar.py
numpy 1.16.2 opticar.py
opencv-python 4.1.0 opticar.py
matplotlib 3.0.3 opticar.py
tqdm 4.45.0 opticar.py
Pillow 5.4.1 sortbg.py
PyAutoGUI 0.9.50 auto.py, autoinfo.py
pynput 1.6.8 autoinfo.py

To install all of the above:

$ pip install tensorflow numpy opencv-python matplotlib tqdm pillow pyautogui pynput

How to run

To train the model, download the dataset, and pass the dataset-directory and number of epochs to train to the script like so:

# example: $ python3 opticar.py train "cars/images/" 30
$ python3 opticar.py train DATASET-DIR NUM-EPOCHS

To use a trained model to predict prices, place the model's weight file model.h5 into the same directory as the script, and place all of the car images that you want to run the model on in another directory. Then run the script like so:

# example: $ python3 opticar.py "cars/images/"
$ python3 opticar.py IMAGES-DIR