im not a big fan of stocks but how does stuff like this happen?

swapped out some layers in the cnn, now it's picking up tumor patterns more clearly. glioma and meningioma detection looking way better. still gotta test on more real-world data tho github.com/qqxnsx/bs-ai/t…

decided to use transfer learning with vgg16. saves time and gives the model a solid starting point instead of training from scratch. accuracy already looking better github.com/qqxnsx/bs-ai/t…

added data augmentation so the model doesn’t just memorize stuff. rotating, flipping, and shifting mri scans should make it handle real-world cases better. training again now github.com/qqxnsx/bs-ai/t…

!

built a way to actually see what the model is focusing on with grad-cam. now we can tell if it's looking at the right spots in the mri or just guessing based on weird noise

so far so good ETA < 1week

ran the model on test data and it’s doing decent, but still getting some false positives on no-tumor cases. might need to tweak thresholds or add more samples for balance

running into some familie issues so i will not be working on the project this week stay tuned for more info, while am at it take a look at this beta github.com/qqxnsx/bs-ai/t…

back to work confusion matrix results are in. glioma and pituitary detection are solid but no-tumor cases still a bit off. probably need to rebalance the dataset or adjust loss function github.com/qqxnsx/bs-ai/t…

finally tested on some mri scans outside the dataset. model did alright but struggled with some blurry ones. might need to train it with lower-quality images too so it doesn’t freak out

getting ready for a beta launch. model is solid but not perfect, so now it’s time to get real feedback. if you know someone in the medical field who wants to test it, hit me up

its officially ready... will talk more about it soon

today is officially the day im so excited

Wait for my official announcement on the supply lock