autodistill-metaclip VS aphantasia

I have been playing with MetaCLIP this afternoon and made https://github.com/autodistill/autodistill-metaclip as a pip installable version. The Facebook repository has some guidance but you have to pull the weights yourself, save them, etc.

My inference function (model.predict("image.png")) return an sv.Classifications object that you can load into supervision for processing (i.e. get top k) [1].

The paper [2] notes the following in terms of performance:

> In Table 4, we observe that MetaCLIP outperforms OpenAI CLIP on ImageNet and average accuracy across 26 tasks, for 3 model scales. With 400 million training data points on ViT-B/32, MetaCLIP outperforms CLIP by +2.1% on ImageNet and by +1.6% on average. On ViT-B/16, MetaCLIP outperforms CLIP by +2.5% on ImageNet and by +1.5% on average. On ViT-L/14, MetaCLIP outperforms CLIP by +0.7% on ImageNet and by +1.4% on average across the 26 tasks.

[1] https://github.com/autodistill/autodistill-metaclip

A new music video I made off of my latest release "Junglehaus" I used the Aphantasia library from eps696 (https://github.com/eps696/aphantasia) by feeding it the lyrics from the song and then editing together the best generations.

(Added Mar. 1, 2021) Aphantasia.ipynb - Colaboratory by eps696. Uses FFT (Fast Fourier Transform) from Lucent/Lucid to generate images. GitHub. Twitter reference. Example #1. Example #2.

Made with Illustrip by Vadim Epstein (https://github.com/eps696/aphantasia) and FL Studio for the background ambience

GPT-J - from the wizards at Eleuther.ai, via HuggingFace. - https://huggingface.co/EleutherAI/gpt-j-6B Aphantasia - from vadim epstein (eps696) - https://github.com/eps696/aphantasia

Over the weekend i've been toying around with the amazing Aphantasia, using quotes about the months of the year as prompts, this is definitely my favorite of the whole set.