Assignment $3$

Due $24$ May by $23$:$59$ Points $42$ Submitting an external tool Available after $6$ May at $0$:$00$

In this assignment, you will implement different deep$-$learning networks to classify images of hands in poses that correspond to letters in American Sign Language. The dataset is contained in the assignment zip file, along with some images and a text file describing the dataset. It is similar in many ways to other MNIST datasets.

The American Sign Language letter database of hand gestures represents a multi$-$class problem with $24$ classes of letters $($excluding J and Z$,$ which require motion$).$ The dataset format is patterned to match closely with the classic MNIST. Each training and test case represents a label $(0-25)$ as a one$-$to$-$one map for each alphabetic letter A$-$Z $($and no cases for $9=$J or $25=$Z because of gesture motions$).$ The training data $(27,455$ cases$)$ and test data $(7172$ cases$)$ are approximately half the size of the standard MNIST but otherwise similar with a header row of labels, which represents a single $28$x$28$ pixel image with grayscale values between $0-255.$

Downloads:

Train and test datasetsDownload Train and test datasets

new sampleDownload new sample

This zip file Download This zip filecontains a Jupyter Notebook template.

IMPORTANT

Train all the models locally on your own machine. No model training should occur on Gradescope $($GS$).$

After completing the training, upload the trained models' h$5$ files and their training histories along with your notebook to GS$.$

best$\_$dnn$\_$bn$\_$model.keras

best$\_$dnn$\_$bn$\_$perf$\_$model.keras

best$\_$dnn$\_$selu$\_$model.keras

best$\_$mobilenet$\_$model.keras

history$1$

history$2$

history$1\_$perf

historymb

To avoid any confusion and poor training on GS$,$ please remember to comment out the training code in your notebook before uploading it to GS$.$