As I did 2 years ago on my DrummerNet paper, I’m open-sourcing the reviews my submission received. I did it back then, and I’m doing it again now, since when I had no paper in ISMIR, I was always very curious about how ISMIR review is being done. By not making this information available, I…More
slightly better research code – avoid hard-coded values
Imagine you need to crop the first 10 second of a waveform. This can be improved by like this. Of course it does the same thing. But this is better because.. Now you know the meaning of this magic number 160000 . And this means that.. Now ANYONE would know the meaning of 160000. Because…More
Tensorflow2 Keras – Custom loss function and metric classes for multi task learning
It is well known that we can use a masking loss for missing-label data, which happens a lot in multi-task learning (example). But how about metrics? Without a similar modification, the keras.metric classes and functions would get you some numbers, but they won’t be quite accurate. No worries though, they can be modified as in…More
Kapre doc → kapre.readthedocs.io
Recently, I put some effort to improve it. Now it supports Tensorflow 2.0. Please enjoy! https://kapre.readthedocs.io/en/latest/More
Some choices I’ve made and why
Only occasionally though, I’ve been asked those classic questions like “So how did you start your career?”, “What motivated you to start a PhD course?”, etc., and somehow I ended up promising that I’ll write a post about it. So, here we go. Disclaimer: I’ll be only straightforward, simple, and dumb. Bachelor: EE My tutor…More
Q&A: How to transcribe rap songs
… I want to understand what they are rapping about … I want to ask if it is indeed possible to transcribe rap songs? I have vocals extracted from the songs and tried to use Google speech2text API for it but the results look very random and bad. I am given the impression that transcribing…More
ICASSP 2020 papers and summaries
Let me reuse my tweets 🙂 https://t.co/ZABBXEDS1c "Improving Universal Sound Separation Using Sound Classification". Used a pre-trained net to extract an embedding that conditions a separation model. Nice work! Turned out it's the same first author (@ETzinis) of the paper above. — Keunwoo Choi (@keunwoochoi) May 18, 2020 https://t.co/hgfCBMnRSU The structure of separate formant mask…More
ICLR 2020 – Invited talk by Michael Jordan
Invited talk by Michael Jordan Michael Jordan first took some time to talk about his idea about the next generation ML – which would be followed by existing ML applications in the real world. He calls it “Markets” – beyond backend (fraud detection, search), human side (RecSys, social media), pattern recognition (now — speech recognition,…More
Tensorflow – parse tfrecords, tf.io.VarLenFeature(tf.string), etc
Sometimes your labels might be something like [‘text1’, ‘test2′] for each example. Say, it’s an image dataset and there’s a label for multiple objects existing in an image. Creating dataset Tensorflow documentation shows that we’ll need to use tf.train.BytesList(). def _bytes_feature(values): values = [v for v in values if v is not None] # i…More
Once seemingly memory leakage with Tensorflow
I had this OOM problem which seemed pretty random before I compared the memory usage per second and Tensorflow log. The problem was not a memory leak. It died at the transition of evaluation and training. I’m quite sure this is happening because the dataloader from the previous set (e.g., training) is still there in…More
