Recognition of digits from handwriting samples
یازده ماه پیش منتشر شده
تعداد بازدید: 168
کد پروژه: 465082
شرح پروژه
We want to build a system for the automatic recognition of digits (0-9)
starting from handwriting samples (8×8 images). In particular, we want to compare the
performance of different types of classifiers. We also want to determine which pairs of
digits are more difficult to discriminate and why.
Method:
- Display the dataset using the PCA.
- Implement two multinomial classifiers (Naïve Bayesian and logistic). Compare the
confusion matrices of the resulting classification process, both for the training set
and for the test set.
- Analyze the performance of each classifier (in terms of % of correct decisions,
AUC) and decide which classifier is best.
- Discuss (for the most problematic pairs of digits from a classification point of view)
which image features are most important for discrimination.
Bonus: Build an additional classifier of your choice and compare its performance with
the previous ones.
این پروژه شامل 1 فایل مهم است، لطفا قبل از ارسال پیشنهاد حتما نسبت به بررسی این فایل اقدام فرمایید.
مهارت ها و تخصص های مورد نیاز
بودجه
750,000 تومان تا 5,000,000 تومان
مهلت برای انجام
20روز
وضعیت مناقصه
انجام شده
درباره کارفرما
عضویت یازده ماه پیش
