}})
+ }})
+ }})
+ }})
+ 
+ An Idea in the Making
+August, 15
+SkinAI started off as a simple idea to identify cysts on one's skin through uploading images. This was our choice since it combined computer vision with a simple product that we could make in a short period of time.
+
+ Collecting Data
+August, 16
+After we chose the final product, we started on collecting data for the model. Each class of skin condition we made had to have about 100 unique images. Then we uploaded them all to Roboflow in order for our team members to annotate and draw bounding boxes on each image for the AI model to learn from. Finally, in order to increase the final size of our data set, we decided to use Roboflow again,to augment each image with rotation, blur, and noise.
+
+ Training the Model
+August, 18
+With our dataset complete, our team started to train the CV model with YOLOv7 which is a fast and accurate object detection software, however we still faced some difficulties.
+
+ Improving the Model
+August, 22
+After our model finished training, the back-end team noticed the extremely low accuracy in detecting the correct skin condition. To try and fix this, they removed rashes for its similarity to certain acne, Ichthyosis because the model kept confusing it for healthy skin, and then added melanoma because we needed another class after removing two, moreover it looking very distinct.We also switched to Yolov5 for its quicker and better results for our purposes +
+
+ Creating the Website
+August, 22
+While the back-end team worked on the code, the front-end team worked on the UI. We used an emotion detector website as our template to work off of. We then worked on getting everything to fit the theme of our logo. Finally we got everything to look a bit more high quality by adding a fading navigation bar as well as adding a nice looking timeline.
+}})
+ }})
+ }})
+ 
+ 
+ }})
+