Need string to int mapping for species labels used in the phase 2 pre-trained model - Githubissues

Evolving-AI-Lab / deep_learning_for_camera_trap_images

This repo contains code+pre-trained models for extracting information from camera-trap images. The pre-trained models have been trained on the Snapshot Serengeti dataset.

MIT License

113 stars 52 forks source link

Need string to int mapping for species labels used in the phase 2 pre-trained model #6

Open alekswithakayy opened 5 years ago

alekswithakayy commented 5 years ago

I can't seem to find how the species labels were mapped from strings to integers for the phase 2 pre-trained model. Was a specific mapping file used or were the animal names sorted and then numbered somehow?

arashno commented 5 years ago

Hi, Animal names sorted and then they numbered from 0 to 47. Here is the mapping: 0|aardvark 1|aardwolf 2|baboon 3|batEaredFox 4|buffalo 5|bushbuck 6|caracal 7|cheetah 8|civet 9|dikDik 10|eland 11|elephant 12|gazelleGrants 13|gazelleThomsons 14|genet 15|giraffe 16|guineaFowl 17|hare 18|hartebeest 19|hippopotamus 20|honeyBadger 21|human 22|hyenaSpotted 23|hyenaStriped 24|impala 25|jackal 26|koriBustard 27|leopard 28|lionFemale 29|lionMale 30|mongoose 31|ostrich 32|otherBird 33|porcupine 34|reedbuck 35|reptiles 36|rhinoceros 37|rodents 38|secretaryBird 39|serval 40|topi 41|vervetMonkey 42|warthog 43|waterbuck 44|wildcat 45|wildebeest 46|zebra 47|zorilla

wei201708 commented 2 years ago

I try an identification of elephant photo using the phase2 with pre-trained model and I get the label 21? Is the label list above right?