CSSE463: Image Recognition Day 20 Project Teams: Stabilizer: Yuzong G, Jackie Z, Tony Y, Yolande X Champion Annotator: Thomas D, Ben K, Tim M, Dan S Automobiles : Larry G, Nicole M, Aaron M, Jarvis Z Math Handwriting: Josh G, Dustin G, Joe M Deep Learning: Nathan B, Matthew G, Jared H, Matthew P  ???: Josh L, Max M, Runzhi Y, Chi Z

Term project next steps From the Lit Review specification. Goal: Review what others have done. Don’t re-invent the wheel. Read papers! Summarize papers Due next Sunday night

CSSE463: Image Recognition Day 20 Today: Lab for sunset detector Next week: Monday: Sunset detector worktime Tuesday: Midterm Exam Thursday: k-means clustering (sunset due 11:00 pm) Friday: lab 6 (k-means)

Exam prep Bright blue roadmap sheet Exam review slides (courtesy reminder)

Last words on neural nets/SVM

Common model of learning machines Statistical Learning (svmtrain) Labeled Training Images Normalize in same way Extract Features (color, texture) Summary Test Image Extract Features (color, texture) Classifier (svmfwd) Label

Sunset Process I suggest writing as you go Loop over 4 folders of images Extract features Normalize Split into train and test and label Save Loop over kernel params Train Test Record accuracy, #sup vec For SVM with param giving best accuracy, Generate ROC curve Find good images Do extension I suggest writing as you go