search

ritwik12 / Celestial-bodies-detection

TensorFlow Image Classifier that can be used to classify whether an image is of a Planet (Earth, Mercury, Mars, etc), Galaxy (Spiral, Elliptical, Irregular), Satellites, Comets, Etc.
https://celestial-bodies-detection.herokuapp.com
GNU General Public License v3.0
84 stars 60 forks source link

Classifying Celestial-bodies #38 #60

Open krkakade opened 4 years ago

krkakade commented 4 years ago

I have created a new branch in which made different commits for different changes. I hope this helps.

ritwik12 commented 4 years ago

@krkakade Where did you find the training data and testing data i.e. images for 'stars', 'nebula', 'supernova', 'cluster_of_galaxies' If you have scrapped these from google then they are all wrong.

If you notice closely then all these catagories will have almost same images and we can't just give the images to our model and train them. We were able to use generic model for classification of planets and types of galaxies because they look different from each other and easy for the model to train. Whereas Nebula, Supernova and cluster_of_galaxies are almost equal.

Better if you create separate PR for each category so that we will be able to check properly.

krkakade commented 4 years ago

According to you, which sources I shoud consider to get data from?

ritwik12 commented 4 years ago

@krkakade I am not sure, we should get from somewhere where it is genuine. Such as satellite images from space authorities through their websites or manually validating images when downloading from google or random place.

I downloaded few from Nasa and other places and manually verified the ones downloaded from google images to be sure.