Trained with Leco score (<a target="_blank" rel="ugc" href="https://civarchive.com/articles/5416">https://civarchive.com/articles/5416</a>)The dataset is 5k images that have the most reactions on civitai (positive label) versus 10k images that are randomly chosen (negative label).Weight 0.2~1.5Remarks:<ul><li>1.5 is not the best weight, pick a good seed, and then try weights from 0.3 to 1.5 (they all give different results)</li><li>trained on AnyLora with LecoScoreV2 for 3000 steps along the PartiPrompts</li><li>training a score on this dataset is unstable (because the difference between very good images and ok images is very small, the network can pickup different characteristics of trending images each time)</li><li>the criterion (trending images) is bad. The main reason for this is that it can indicate a large number of followers or a very 'mainstream' theme for the image. I would prefer the number of positive reactions per view (the eye with a number when you click on an image), but this would be very noisy as well.</li><li>dataset is included (if you want to train something on it)</li><li>example images are sometimes rendered within the Lykon universe (this was trained on AnyLora, then use Dreamshaper, Fast Negative, Easy Negative), I had to remove the Lykon negatives for other models</li><li>-&gt; alpha version</li></ul>Edit: <ul><li>something has been bothering me about this model, mostly its 'mainstream aesthetics', lots of different colors, bokeh, barbie doll anatomy. In other words, complete lack of artistic flair (gritty but elating, ugly-cute aka gurokawai, charming appearance instead of just perfect body).</li><li>-&gt; I attended a lecture which included a slide with 'people's ingrained image types'. It showed pictures of food (stock images) versus rotten food, and said we are biologically and evolutionarily wired to prefer the pictures of fresh food, so that we will eat non toxic food (I really disliked the soulless stock pictures though, it felt like junk food for the eyes). It is possible that any model like this one will, based on the average judgement of thousands, just pick up images that show low level psychological biases (lots of colors for freshness, high hip to waist ratio for women, uncanny valley exclusion, etc...).</li><li>it is therefore possible that this Lora, as well as aesthetic image scores, are not able to differentiate good drawing skills (good art) - versus - aesthetically pleasing for human brain evolutionary reasons (ingrained pleasant artistic content). There is only one 'aesthetic score'.</li></ul><ul><li>Related, found on the Internets:</li></ul><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/738ce6b7-1497-4d5c-8102-fb3a8786d1ec/width=525/738ce6b7-1497-4d5c-8102-fb3a8786d1ec.jpeg" />

Files with SHA256: 6feafc1130b617def86df695b4ff4cd529736a97970b4b7e21e694cd2cd5cbc9