Inspiration
It is very time consuming, intensive and inconvenient for online sellers selling a large variety of clothing to write out captions for each article of clothing. From personal experience, when moving house and clearing out the wardrobe, there is a lot of clothing and other items that we want to sell and it is very time consuming to manually caption everything.
What it does
Auto captioning and auto tagging, through our model identifying different types of clothes from the uploaded images to save sellers’ time from typing in their captions and tagging their items manually. This is especially useful in large scale manufacturing and selling.
How we built it
Using Python, Google Colab and OpenCV, we built and trained a model using the dataset provided to us by BuildingBloCS to identify and classify various articles of clothing.
Challenges we ran into
At first we were unsure of what theme to choose as none of our initial ideas fit the criteria of being innovative, ingenious and impactful, but eventually after much deliberation, we settled on this final idea. Another challenge was also that the model took a long time to train, and it was especially time-consuming when it was inaccurate and had to be retrained. This was very frustrating and we had to discuss how long we would train it for and then retrain it such that it worked smoothly. This definitely took a lot of time and was quite difficult to navigate.
Accomplishments that we're proud of
The model works well, and it can be used to classify articles of clothing under their types like pants, shirts, skirts, etc.
What we learned
How to use OpenCV to train a model to classify different articles of clothing by labelling them as what type of clothing they are, shirt, pants, hat, dress, etc.
What's next for Group I2
Recommend competitive prices by searching our website for prices of similar items so users can save money Improving our colour recognition system via incorporating object identification Finetuning our model further in general
Model:
https://drive.google.com/drive/folders/1RA2zeM4Ycf2TMK5iIw1yz23r6J3Bkfsx
Log in or sign up for Devpost to join the conversation.