Any idea where to find a database of birds? I've been kicking around with the idea of using transfer learning and then a database of bird images to fine tune a classifier to tell people what a bird they take a picture of is (EDIT, similar approach: [0]).
Would help with research. I know some people in finland that had participated in a crowdsourced census of birdlife in their area and thought it could be easier with an app that captures bird images and automatically classifies them (tells the user, I've often wondered what kind of bird something I've seen is) and provides the data to researchers.
I asked a friend of mine who works at a bird life charity if he knew of any data sets, but he didn't.
I was looking into this for identifying other animals.
iNaturalist has a community that adds observations and identifies species. They have an API as well as a dataset for research https://www.inaturalist.org/pages/developers
I'd love to see some sample code. I've tried something similar to detect the postman in my security camera stream, but failed over lots of small technical details.
Does anyone know a good tutorial on reinforcement learning that describes a practical application like this?
I wanted to try out different types of final layers (SVM, Logistic Reg.) so the direct example didn't suit very well for that kind of testing out. For starters that good though.
I will probably publish in some time (when I'm not so busy doing other projects) some code samples too but until that I can give you tips.
You should start inspecting how the pre trained models (Inception v3/v4) works, what kind of layers they have and then decide what layers you want to use and what not. In case of Tensorflow, the tensorboard is very good tool for inspecting the model inner layers.
The first lesson in the Fast.ai deep learning course covers exactly this (retraining a pre-trained network, in this case VGG, to distinguish between cats and dogs). There's a video and an accompanying ipython notebook with sample code, using Keras.
You are using a “ready” product and so everyone gets the updates to model but your own modifications for the model are impossible with current providers.
I think this may not be true about Azure. In Build conference this year, they demoed cognitive service APIs where one can use transfer learning techniques, and train models and use them.
That sounds interesting, it's funny if you really can use their own models as base and do that. For the platform sake, Google offers also SaaS where you can train and evaluate your own models but then the base model is something you have to provide yourself
EDIT: I tried to google that up but couldn't find anything. Could you provide a link for that
That is great and will definetly help with problems where your task isn't just to recognize cats and dogs the only downside is that you are giving your data away and it will also help your competitors.
20 comments
[ 3.6 ms ] story [ 30.0 ms ] threadTwitter account is located here [1].
[0] https://wwf.fi/en/norppalive/
[1] https://twitter.com/NorppaVahti
In case you guys have ideas for other similar ML/AI projects - we could take a look!
(shameless plug, I'm a partner at Emblica.fi)
Would help with research. I know some people in finland that had participated in a crowdsourced census of birdlife in their area and thought it could be easier with an app that captures bird images and automatically classifies them (tells the user, I've often wondered what kind of bird something I've seen is) and provides the data to researchers.
I asked a friend of mine who works at a bird life charity if he knew of any data sets, but he didn't.
[0] https://jeffxtang.github.io/deep/learning,/tensorflow,/mobil...
But it's North American birds, not sure if you specifically want to cover birds that show up in Finland or if that was just your inspiration.
There are a couple apps that do what you're talking about, but they're both North American birds only as well.
http://www.birdsnap.com/ (I was involved in this one)
http://merlin.allaboutbirds.org/
Does anyone know a good tutorial on reinforcement learning that describes a practical application like this?
Follow this to retrain the Inception network: https://www.tensorflow.org/tutorials/image_retraining
Follow this to run an image through your model for classification: https://www.tensorflow.org/tutorials/image_recognition
I will probably publish in some time (when I'm not so busy doing other projects) some code samples too but until that I can give you tips.
You should start inspecting how the pre trained models (Inception v3/v4) works, what kind of layers they have and then decide what layers you want to use and what not. In case of Tensorflow, the tensorboard is very good tool for inspecting the model inner layers.
If you want to get started even more easier you should probably take a look into Tensorflow Slim models (https://github.com/tensorflow/models/tree/master/slim)
There is quite beginner friendly instructions for simple fine tuning of the models and it should take you pretty far.
http://course.fast.ai/lessons/lesson1.html
You are using a “ready” product and so everyone gets the updates to model but your own modifications for the model are impossible with current providers.
I think this may not be true about Azure. In Build conference this year, they demoed cognitive service APIs where one can use transfer learning techniques, and train models and use them.
EDIT: I tried to google that up but couldn't find anything. Could you provide a link for that
https://azure.microsoft.com/en-us/services/cognitive-service...
That is great and will definetly help with problems where your task isn't just to recognize cats and dogs the only downside is that you are giving your data away and it will also help your competitors.