I built a neural network that guesses the genre of a music based on data
scrapped on Spotify!
Report Bug
·
Request Feature
prediction: punk | real result: punk
prediction: country | real result: country
good/total: 28/30 - 93%
- Install dependancies
pip install -r requirements.txt
- Run
python train.py
- Open the dashboard
tensorboard --logdir=logs
- Install dependancies
pip install -r requirements.txt
-
Change
NUMBER_VALUESintest.py -
Run!
python test.py
prediction: pop | real result: pop
prediction: rock | real result: rock
prediction: pop | real result: pop
prediction: edm | real result: edm
prediction: jazz | real result: jazz
prediction: electro | real result: electro
prediction: pop | real result: pop
prediction: rock | real result: rock
prediction: salsa | real result: salsa
prediction: chanson | real result: chanson
prediction: chanson | real result: hip hop
prediction: rock | real result: rock
prediction: hip hop | real result: hip hop
prediction: dance | real result: dance
prediction: pop | real result: pop
prediction: hip hop | real result: soul
prediction: jazz | real result: jazz
prediction: pop | real result: pop
prediction: hip hop | real result: hip hop
prediction: hip hop | real result: hip hop
prediction: pop | real result: pop
prediction: edm | real result: edm
prediction: dance | real result: dance
prediction: rock | real result: rock
prediction: disco | real result: disco
prediction: rock | real result: rock
prediction: country | real result: country
prediction: blues | real result: blues
prediction: punk | real result: punk
prediction: country | real result: country
good/total: 28/30 - 93%
- with a long training acc should be ok, but test dataset has low acc. maybe it's overtraining...
- change the model (layers, categorical to binary?)
- drop some col on the dataset that are useless
- check normalization