Alternate Python bindings for the Open Asset Import Library (ASSIMP)

Overview

Impasse

Python Test Status codecov

A simple Python wrapper for assimp using cffi to access the library. Requires Python >= 3.7.

It's a fork of PyAssimp, Assimp's official Python port. In contrast to PyAssimp, it strictly targets modern Python 3 and provides type hints. It also aims to allow mutating scenes before exporting by having all wrapper classes operate directly on the underlying C data structures.

Usage

Complete example: 3D viewer

impasse comes with a simple 3D viewer that shows how to load and display a 3D model using a shader-based OpenGL pipeline.

Screenshot

To use it:

python ./scripts/3d_viewer.py <path to your model>

You can use this code as starting point in your applications.

Writing your own code

To get started with impasse, examine the simpler sample.py script in scripts/, which illustrates the basic usage. All Assimp data structures are wrapped using ctypes. All the data+length fields in Assimp's data structures (such as aiMesh::mNumVertices, aiMesh::mVertices) are replaced by list-like wrapper classes, so you can call len() on them to get their respective size and access members using [].

For example, to load a file named hello.3ds and print the first vertex of the first mesh, you would do (proper error handling substituted by assertions ...):

from impasse import load

scene = load('hello.3ds')

assert len(scene.meshes)
mesh = scene.meshes[0]

assert len(mesh.vertices)
print(mesh.vertices[0])

Another example to list the 'top nodes' in a scene:

from impasse import load

scene = load('hello.3ds')
for c in scene.root_node.children:
    print(str(c))

All of assimp's coordinate classes are returned as NumPy arrays, so you can work with them using library for 3d math that handles NumPy arrays. Using transforms.py to modify the scene:

import math

import numpy
import transformations
import impasse

# assimp returns an immutable scene, we have to copy it if we want to change it
scene = impasse.load('hello.3ds').copy_mutable()
transform = scene.root_node.transformation
# Rotate the root node's transform by 180 deg on X
transform = numpy.dot(transformations.rotation_matrix(math.pi, (1, 0, 0)), transform)
scene.root_node.transformation = transform
impasse.export(scene, 'whatever.obj', 'obj')

Installing

Install impasse by running:

pip install impasse

or, if you want to install from the source directory:

pip install -e .

Impasse requires an assimp dynamic library (DLL on Windows, .so on linux, .dynlib on macOS) in order to work. The default search directories are:

  • the current directory
  • on linux additionally: /usr/lib, /usr/local/lib, /usr/lib/ -linux-gnu

To build that library, refer to the Assimp master INSTALL instructions. To look in more places, edit ./impasse/helper.py. There's an additional_dirs list waiting for your entries.

Progress

All features present in PyAssimp are now present in Assimp (plus a few more!) Since the API largely mirrors PyAssimp's, most existing code should work in Impasse with minor changes.

Note that Impasse is not complete. Many assimp features are still missing, mostly around mutating scenes. Notably, anything that would require a new or delete in assimp's C++ API is not supported.

Performance

Impasse tries to avoid unnecessary copies or conversions of data owned by C, and most classes are just thin layers around the underlying CFFI structs. NumPy arrays that directly map to the underlying structs' memory are used for the coordinate structs like Matrix4x4 and Vector3D.

Testing with a similar quicktest.py script against assimp's test model directory:

Impasse

** Loaded 169 models, got controlled errors for 28 files, 0 uncontrolled

real	0m1.460s
user	0m1.676s
sys	0m0.571s

PyAssimp

** Loaded 165 models, got controlled errors for 28 files, 4 uncontrolled

real	0m7.607s
user	0m7.746s
sys	0m0.579s
Comments
  • Nicer way of referring to material property keys

    Nicer way of referring to material property keys

        ('$mat.twosided', 0): [0]
        ('$mat.refracti', 0): [1.]
        ('$mat.bumpscaling', 0): [1.]
        ('$clr.specular', 0): [1. 1. 1.]
    

    Those keys are in the data returned by assimp, but I'm not sure if the prefix is that meaningful. PyAssimp strips off everything before the first .. If they're meaningful or there are collisions without those prefixes, we should keep a string enum of common ones that's easier to refer to.

    enhancement 
    opened by SaladDais 3
  • Rebase on top of original assimp sources to keep PyAssimp's commit history intact

    Rebase on top of original assimp sources to keep PyAssimp's commit history intact

    Now that I think of it, it's preferable to keep the commit history for assimp even if most of it's unrelated to PyAssimp. Make a commit deleting all assimp sources and moving PyAssimp to the repo root, then rebase impasse on top of that.

    Rewriting history isn't nice since I've already based my sources on top of a squashed commit, but impact should be mininal since nobody's using this yet.

    opened by SaladDais 1
  • Make mapping classes mix-ins

    Make mapping classes mix-ins

    Other than name collisions with keys / values, I don't think there's any reason to keep the explicit as_mapping() method rather than just putting them on a mixin for the SerializeableStructs. Can define manual renames for any of those.

    This'd give us a closer API to what PyAssimp already has

    enhancement 
    opened by SaladDais 1
  • Add accessors for mesh and texture that return instances rather than indexes

    Add accessors for mesh and texture that return instances rather than indexes

    All of the struct wrapper have a _scene member that should allow them to transparently look up the texture / material by the index in the attr they wrap.

    enhancement 
    opened by SaladDais 1
  • Get scene mutability working

    Get scene mutability working

    Per the C api you have to create a copy of the scene to get a non-const version: https://github.com/assimp/assimp/blob/master/include/assimp/cexport.h#L109-L123 . Right now all the structs we return have wrappers enforcing the const-ness of scene returned from the the aiImportFile() function. Need to expose the copying functions so people can get a scene they can modify before export.

    enhancement 
    opened by SaladDais 1
  • Package assimp shared library so assimp isn't required to be on the system

    Package assimp shared library so assimp isn't required to be on the system

    Would be nice so people don't have to do some weird conda thing if they want to install the package including assimp.

    I think it should be enough to have osx aarch64, osx x64, linux x64 and windows x64 packages. We don't need to do python version-specific builds, since we'll just be building the assimp library itself and binding against it with cffi. It should be provided in an optional, separate package with the shared library in the package data for the given platform's build.

    enhancement 
    opened by SaladDais 0
  • Add ability to alloc new structs, append to sequences

    Add ability to alloc new structs, append to sequences

    This is tricky since assimp makes liberal use of new[] and delete[] on types that have non-trivial destructors (like aiTexture.) malloc()ing and free()ing those is technically possible but inadvisable, and requires knowledge of the platform / compiler's C++ ABI.

    Likewise, assimp's C API doesn't appear to expose a wrapper around those new[] and delete[] calls. Adding functions to assimp to do those would be the least nasal demon-y approach but wouldn't be available in stable distros' libs.

    A third approach would be to have Scenes keep track of the original values of mutated ptrs, then put everything back in place when the GC happens so anything that'd been internally alloc'd with new could be deleted by assimp. Would also have to keep a list of things we'd malloc()d in our own code to manually free(). Can see a lot of nasty corner cases with this one.

    enhancement help wanted 
    opened by SaladDais 0
Releases(v5.2.0)
Owner
Salad Dais
Code as craft
Salad Dais
Fixes 500+ mislabeled MURA images

In this repository, new csv files are provided that fixes 500+ mislabeled MURA x-rays for all categories. The mislabeled x-rays mainly had hardware in them. This project only fixes the false negative

Pieter Zeilstra 4 May 18, 2022
Simple Python image processing & automatization project for a simple web based game

What is this? Simple Python image processing & automatization project for a simple web based game Made using only Github Copilot (except the color and

SGeri 2 Aug 15, 2022
A Toolbox for Image Feature Matching and Evaluations

This is a toolbox repository to help evaluate various methods that perform image matching from a pair of images.

Qunjie Zhou 342 Dec 29, 2022
Fast Image Retrieval is an open source image retrieval framework

Fast Image Retrieval is an open source image retrieval framework release by Center of Image and Signal Processing Lab (CISiP Lab), Universiti Malaya. This framework implements most of the major binar

CISiP Lab 39 Nov 25, 2022
Kimimaro: Skeletonize Densely Labeled Images

Kimimaro: Skeletonize Densely Labeled Images # Produce SWC files from volumetric images. kimimaro forge labels.npy --progress # writes to ./kimimaro_o

92 Dec 17, 2022
Python-based tools for document analysis and OCR

ocropy OCRopus is a collection of document analysis programs, not a turn-key OCR system. In order to apply it to your documents, you may need to do so

OCRopus 3.2k Jan 04, 2023
Program for analyzing shadows from Cassini images

Moons: An Analysis Module for Vicar Files General This packages/program was created for my bachelor's thesis for the Astronomy department at Universit

Joni 1 Jul 16, 2021
Demo of using Auto Encoder for Image Denoising

Demo of using Auto Encoder for Image Denoising

2 Aug 04, 2022
Optimize/Compress images using python

Image Optimization Using Python steps to run the script run the command to install the required libraries pip install -r requirements.txt create a dir

Shekhar Gupta 1 Oct 15, 2021
QSIprep: Preprocessing and analysis of q-space images

QSIprep: Preprocessing and analysis of q-space images Full documentation at https://qsiprep.readthedocs.io About qsiprep configures pipelines for proc

Lifespan Informatics and Neuroimaging Center 88 Dec 15, 2022
Advance Image Steganography

Chaya Advance Image Steganography Using LSB-LPS + AES-256-GCM + FLIF Right To Privacy! United Nations Declaration of Human Rights (UDHR) 1948, Article

XeroHack 113 Dec 02, 2022
Python pygame project that turns your images to matrix rain

Matrix-Rain-An-Image This project implements the classic Matrix digital rain effect in python with pygame to build up an image provided with multiple

7 Dec 11, 2022
The following program is used to swap the faces from two images.

Face-Swapping The following program is used to swap the faces from two images. In today's world deep fake technology has become really popular . As a

1 Jan 19, 2022
Convert photos to paintings with python

Convert-photos-to-paintings Before the changes After the changes Before the changes After the changes This code is written in the Python programming l

Amir Hussein Sharifnezhad 3 May 31, 2022
This app finds duplicate to near duplicate images by generating a hash value for each image stored with a specialized data structure called VP-Tree which makes searching an image on a dataset of 100Ks almost instantanious

Offline Reverse Image Search Overview This app finds duplicate to near duplicate images by generating a hash value for each image stored with a specia

53 Nov 15, 2022
LGVL helper script to batch and convert with lvgl offline image converter

script to batch and convert with lvgl offline image converter

Yohann 1 Oct 05, 2022
Combinatorial image generator for generative NFT art.

ImageGen Stitches multiple image layers together into one image. Run usage: stitch.py [-h] backgrounds_dir dinos_dir traits_dir texture_file

Dinosols NFT 19 Sep 16, 2022
Console images in 48 colors, 216 colors and full rgb

console_images Console images in 48 colors, 216 colors and full rgb Full RGB 216 colors 48 colors If it does not work maybe you should change color_fu

Урядов Алексей 5 Oct 11, 2022
Ascify-Art - An easy to use, GUI based and user-friendly colored ASCII art generator from images!

Ascify-Art This is a python based colored ASCII art generator for free! How to Install? You can download and use the python version if you want, modul

Akash Bora 14 Dec 31, 2022
A linear stairs generation add-on for Blender

Linear Stairs Generator Table of Contents Installation Usage Screenshots Important Notes Requirements Blender 3.0 or newer. Installation: Download a z

Elhanan Flesch 4 May 17, 2022