La arduino raya hukumar Yana ba da damar dubban dubban ayyuka daban-daban da za a yi, iyaka a zahiri yana cikin tunanin kowane mai yin, ko da yake yana da wasu gazawar jiki, kamar ƙwaƙwalwar ajiya, iya aiki, da sauransu. Duk da haka, akwai ƙarin samfurori da ayyuka don ƙaddamar da damar su da yawa, kamar yadda lamarin yake sabon ƙaddamar da AIfES.
Yanzu, godiya ga wannan aikin halitta ta Fraunhofer IMS na Arduino, wannan buɗaɗɗen allo zai ƙunshi a Tsarin hankali na wucin gadi (AI) wanda aka tsara a cikin C, ta amfani da daidaitattun ɗakunan karatu na GNU GCC. Masu amfani yanzu za su iya ƙara AIfES zuwa aikin su na Arduino kuma su haɗa shi daga mai kula da ɗakin karatu daga IDE don amfani da shi a cikin ci gaban ku, yana ba da damar yin amfani da algorithms na koyon injin ko da a cikin ƙananan microcontrollers kamar allo. Arduino UNO 8-bit.
Wannan zai ba da damar masu haɓakawa su ƙirƙiri ɗimbin na'urori na IoT (Internet of Things) waɗanda ke da 'yancin kai daga gajimare kuma waɗanda za su iya zama masu hankali, kuma tare da mutunta sirrin ku, tunda ana iya aiwatar da ayyuka ta layi ta layi daga allon Arduino ba tare da buƙata ba. don dogara ga ayyuka masu nisa. Bugu da ƙari, an ƙaddamar da aikin AIfES a ƙarƙashin GNU GPLv3 lasisi, don haka yana da cikakkiyar kyauta, kodayake yana ba da izinin lasisin da aka biya don ayyukan kasuwanci.
AIfES yana da kama da juna kuma yana dacewa da Python ML frameworks kamar yadda lamarin yake tare da TensorFlow, Keras ko PyTorch, amma aikin sa ya ɗan ragu. Koyaya, a cikin wannan sigar da aka saki FNN (Feedforward Neural Networks) an riga an tallafawa, ƙari kuma yana ba da damar kunna ayyukan haɗaka kamar ReLu, Sigmoid, ko Softmax. A gefe guda, masu haɓakawa suna aiki don kawo nan gaba kuma aiwatar da ConvNet (Convolutional Neural Networks), wanda ba zai ɗauki lokaci mai tsawo ba don isa.
Wasu kuma an haɗa su horo algorithms na kowa, kamar SGD (Gradient Descent Optimizer) da Adam Optimizer, da sauransu. Ina nufin, don 8-bit MCU, ba shi da kyau ko kaɗan ...