Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

More After Image system enhancements for future project use. Trying out different blending modes, and mimicking a character after image like Mega Man X4's Dash trail. I think the results are pretty neat! #MegaManRS #UnrealEngine #gamedev

14,014 Aufrufe • vor 10 Monaten •via X (Twitter)

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

How to Bring Emojis to Life with AI (Yes, you read that right) 1. Choose the Emoji Wisely Prefer detailed emojis 👨‍🎤 The Man Singer emoji is a good example of this. Face, hair, outfit, mic... The more detailed, the better. 2. Prepare Your Original Image We'll be doing multi-enhancements, so the initial image shouldn't be too large. (I'll explain more about this shortly) I started with a 512x512px image. The smaller the image you start with, the more detail-oriented the enhancer tends to be. I usually use Figma for design and animations. However, you can use any other tool like Canva or Photoshop. Or just download it from somewhere. 3. Upload Your Image Go to Magnific's website and simply drag and drop your image there. 4. Setup the Parameters This is the fun part. You can experiment here to find your own settings or follow the method I used. I start with high creativity and low resemblance values. Depending on how far you want to deviate from the original image, you can increase this contrast. First Enhancing Creativity: 6 / Resemblance: -3 All Parameters Optimized for: Standard Creativity: 6 HDR: 0 (not a fan) Resemblance: -3 Engine: Automatic 5. The Prompt 🪄 Since Magnific uses Stable Diffusion's technology, it uses the same syntax for prompts. But you don't need to write very detailed prompts. Just state what you want to see. The trick here is to use prompt weights. You need to specify the features you want to emphasize like this: Bowie, Portrait photo of a male singer with blue hair, (detailed eyes:1.2), (detailed hair:1.2), (detailed fabric:1.1), (detailed mic:1.1), (realistic skin:1.5), ultra-realistic, 8k You can change the weight values between 1.0 - 2.0 6. Hit that Rainbow Button After a short wait, you'll have a 1024x1024px image. If you like the result, download the image for further enhancements. 7. Reenhancing The system supports up to approximately 4400 x 4400 px. This means we can enhance it two more times. 1x → 512x512px 2x → 1024x1024px 4x → 2048x2048px 8x → 4096x4096px Now, upload the image you downloaded again. This time, we bring the values closer to each other. Second Enhancing Creativity: 4 / Resemblance: -2 Third Enhancing Creativity: 2 / Resemblance: -1 The other values can remain the same. 8. Share what you create! That's it. Now you know everything I know. If you have any questions, don't hesitate to ask. And please share your creations with us. Happy Sunday everyone!

Dogan Ural

225,681 Aufrufe • vor 2 Jahren

[CLIP] by Hand ✍️ The CLIP (Contrastive Language–Image Pre-training) model, a groundbreaking work by OpenAI, redefines the intersection of computer vision and natural language processing. It is the basis of all the multi-modal foundation models we see today. How does CLIP work? Goal: 🟨 Learn a shared embedding space for text and image [1] Given ↳ A mini batch of 3 text-image pairs ↳ OpenAI used 400 million text-image pairs to train its original CLIP model. Process 1st pair: "big table" [2] 🟪 Text → 2 Vectors (3D) ↳ Look up word embedding vectors using word2vec. [3] 🟩 Image → 2 Vectors (4D) ↳ Divide the image into two patches. ↳ Flatten each patch [4] Process other pairs ↳ Repeat [2]-[3] [5] 🟪 Text Encoder & 🟩 Image Encoder ↳ Encode input vectors into feature vectors ↳ Here, both encoders are simple one layer perceptron (linear + ReLU) ↳ In practice, the encoders are usually transformer models. [6] 🟪 🟩 Mean Pooling: 2 → 1 vector ↳ Average 2 feature vectors into a single vector by averaging across the columns ↳ The goal is to have one vector to represent each image or text [7] 🟪 🟩 -> 🟨 Projection ↳ Note that the text and image feature vectors from the encoders have different dimensions (3D vs. 4D). ↳ Use a linear layer to project image and text vectors to a 2D shared embedding space. 🏋️ Contrastive Pre-training 🏋️ [8] Prepare for MatMul ↳ Copy text vectors (T1,T2,T3) ↳ Copy the transpose of image vectors (I1,I2,I3) ↳ They are all in the 2D shared embedding space. [9] 🟦 MatMul ↳ Multiply T and I matrices. ↳ This is equivalent to taking dot product between every pair of image and text vectors. ↳ The purpose is to use dot product to estimate the similarity between a pair of image-text. [10] 🟦 Softmax: e^x ↳ Raise e to the power of the number in each cell ↳ To simplify hand calculation, we approximate e^□ with 3^□. [11] 🟦 Softmax: ∑ ↳ Sum each row for 🟩 image→🟪 text ↳ Sum each column for 🟪 text→ 🟩 image [12] 🟦 Softmax: 1 / sum ↳ Divide each element by the column sum to obtain a similarity matrix for 🟪 text→🟩 image ↳ Divide each element by the row sum to obtain a similarity matrix for 🟩 image→🟪 text [13] 🟥 Loss Gradients ↳ The "Targets" for the similarity matrices are Identity Matrices. ↳ Why? If I and T come from the same pair (i=j), we want the highest value, which is 1, and 0 otherwise. ↳ Apply the simple equation of [Similarity - Target] to compute gradients of for both directions. ↳ Why so simple? Because when Softmax and Cross-Entropy Loss are used together, the math magically works out that way. ↳ These gradients kick off the backpropagation process to update weights and biases of the encoders and projection layers (red borders).

Tom Yeh

67,834 Aufrufe • vor 2 Jahren