scikit-image VS DenseDepth

We will use the Hugging Face transformers and diffusers libraries for inference, FiftyOne for data management and visualization, and scikit-image for evaluation metrics.

Data analysis involves scrutinizing datasets for class imbalances or protected features and understanding their correlations and representations. A classical tool like pandas would be my obvious choice for most of the analysis, and I would use OpenCV or Scikit-Image for image-related tasks.

scikit hog library: https://github.com/scikit-image/scikit-image/blob/main/skimage/feature/_hog.py#L302 , https://github.com/scikit-image/scikit-image/blob/main/skimage/feature/_hoghistogram.pyx

This is a good cv deep learning book with python examples https://www.manning.com/books/deep-learning-for-vision-systems. If you're pretty comfortable with the concepts of traditional image processing this is a good companion to cv2 (so you don't have to reinvent the wheel) https://scikit-image.org/

Also, don't know if you're familiar with Python, but if you need ideas for to implement for future directions : https://scikit-image.org/

There's probably something in scikit-image to do what you want, or close enough to build on.

Scikit-image The Scikit-image library is a collection of image processing algorithms that are designed to be easy to use and understand. It includes algorithms for common tasks like edge detection, feature extraction, and image restoration. If you are just starting out in image processing, then this is a good library to check out!

Python is a general-purpose programming language that provides many image processing libraries for adding image processing capabilities to digital images. Some of the most common image processing libraries in Python are OpenCV, Python Imaging Library (PIL), Scikit-image etc.

Then you would need to find what this plugin does for imshow. First thing you can see is that "interpolation" is not "bicubic" as you used, but "nearest"… but there are other settings here that are responsible for the difference of displays. (it's better that you look at the source code in your environment, as it might be slightly different)

For a long time, the state-of-the-art models for monocular depth estimation such as DORN and DenseDepth were built with convolutional neural networks. Recently, however, both transformer-based models such as DPT and GLPN, and diffusion-based models like Marigold have achieved remarkable results!

Instead of just doing grayscale for depth, you should consider using “monocular depth estimation” that actually tries to reconstruct a depth map from an RGB image. There are a variety of open-source libs available like this one.

After that I have to do the same for Python I think, and then I have to find out how to figure out how to use a library like https://github.com/ialhashim/DenseDepth or https://github.com/nianticlabs/monodepth2 for that GStreamer plugin (or element, still trying to grasp the terminology here)

Others model implemented like github.com/ialhashim/DenseDepth have constraints on input (I think only 4:3 is one of them, if I remember correctly)