Inspiration
We were totally clueless with nothing at all to start. The first few hours were just passing by as we tried to understand the assignment. We were not expecting this at all, and with this very difficult knowledge it was hard to even write the first line of code. We gained inspiration from other websites and switched from VS code to Google Colab. We also had to switch from another module to CNN tensor flow 2.
What it does
It helps to detect images if they are aliens or humans with ease. It is able to identify simple pictures but may not be complex. It definitely helps to identify genuine and clear aliens and humans.
How we built it
We built it using CNN and trained the model with around 200 pictures. It is well built and is working good.
Challenges we ran into
We ran into many difficulties and challenges like fixing hardcore bugs, finding datasets for aliens and humans, and insufficient knowledge about importing libraries the basic knowledge required for the project.
Accomplishments that we're proud of
We are proud of completing the entire assignment as the model can identify images with 80% accuracy which we had no hopes for at the starting of the event. We had no hopes for even completing or finishing the project completely.
What we learned
We learnt how to fix bugs and how to put your code systematically. We also learnt how to test, train AI to detect images. We also learnt how to make datasets and how to use chrome extensions usefully.
What's next for Alienators
In the future we hope to improve our current model so that we have more time so that the model has more accuracy which includes more data to work on. We hope to improve on our mistakes due to limited knowledge.
resources
https://techzizou.com/training-an-ssd-model-for-a-custom-object-using-tensorflow-2-x/ https://www.kaggle.com/datasets/pmigdal/alien-vs-predator-images?resource=download
Built With
- cnn-tensorflow-2
- google-colab
- python
Log in or sign up for Devpost to join the conversation.