Bike Price-Prediction

This is a personal project finished as a 2 member group-project for university. The project is inspired by Stanford CS229: Machine Learning Course, Lecture 1 - Andrew Ng (Autumn 2018), where one instructor mentions a similar project done by students.

1. Web-Crawling of (bicycle-img, offer-price) mapping, 60000k images
2. Cleanup and analysis of data
3. "Proof of concept" ResNet50 & Robust Model using MAPE (Mean Absolute Percentage Error)
4. Visualization using GradCam
5. Interested in the tensorflow model or clean dataset? Please contact me :)

Results: From PowerPoint slides, the actual notebook "BikePricePrediction" is of course more detailed but also contains university requirements (preliminaries etc.):

Below is the full presentation with a high level overview over the project.

Table of Content

• 1. Problem formulation

• 2. Crawling

• 3. Filtering and cleaning

• 3.1. Filtering out non-bicycle images using ResNet(ImageNet) embeddings

• 3.2. Filter out remaining unwated images

• 3.3. Remove duplicate images

• 3.4. Is there bias?

• 3.5. Preprocessing

• 3.5 Mean Image

• 3.5 Data Split

• 4. Loss Function

• 5. Baseline Model

• 5. "Finetuned" Model

• 5. Robust Model

• 5. Recap: "Is there bias in the data?"

• 6. GradCam Feature Map visualization