A small library for fast Wavefront .obj and .mtl files loading. As .obj parser was used this fantastic library https://github.com/syoyo/tinyobjloader

These instructions will help you to install the package and use it.

Using gradle:

implementation 'com.riseapps.marusyaobjloader:marusyaobjloader:1.0.0'

Using maven:

<dependency> <groupId>com.riseapps.marusyaobjloader</groupId> <artifactId>marusyaobjloader</artifactId> <version>1.0.0</version> <type>pom</type> </dependency>

Or download it as a file here

Use this code for the minimum functionality:

import com.riseapps.marusyaobjloader.MarusyaObjLoader; import com.riseapps.marusyaobjloader.MarusyaObjLoaderImpl; import com.riseapps.marusyaobjloader.model.ResultModel; import java.io.File; import java.io.FileNotFoundException; private static final File FILES_PATH = new File(Environment.getExternalStorageDirectory(), "jni_test"); private static final File OBJ = new File(FILES_PATH, "luxury_house_interior.obj"); private static final File MTL = new File(FILES_PATH, "luxury_house_interior.mtl"); public void load() { final MarusyaObjLoader marusyaObjLoader = new MarusyaObjLoaderImpl(); ResultModel resultModel = null; try { resultModel = marusyaObjLoader.load( OBJ, MTL, 1.0f, true); } catch (FileNotFoundException e) { e.printStackTrace(); } } }

The full example of using is located here

ResultModel load(File obj, boolean flipTextureCoordinates) throws FileNotFoundException;

obj - .obj file in a storage

flipTextureCoordinates - reverse a texture or not

ResultModel - an object which contains shapes, materials, error and warning messages

ResultModel load(File obj, File mtl, boolean flipTextureCoordinates) throws FileNotFoundException;

obj - .obj file in a storage

mtl - .mtl file in a storage

flipTextureCoordinates - reverse a texture or not

ResultModel - an object which contains shapes, materials, error and warning messages

ResultModel load(File obj, float normalizeCoefficient, boolean flipTextureCoordinates) throws FileNotFoundException;

obj - .obj file in a storage

normalizeCoefficient - allows normalizing vertices

flipTextureCoordinates - reverse a texture or not

ResultModel - an object which contains shapes, materials, error and warning messages

ResultModel load(File obj, File mtl, float normalizeCoefficient, boolean flipTextureCoordinates) throws FileNotFoundException;

obj - .obj file in a storage

mtl - .mtl file in a storage

normalizeCoefficient - allows normalizing vertices

flipTextureCoordinates - reverse a texture or not

ResultModel - an object which contains shapes, materials, error and warning messages

void enableLog();

You will see a log like this:

D/MarusyaObjLoader: *********************************************************************************** D/MarusyaObjLoader: Start parsing -> obj: /storage/emulated/0/jni_test/luxury_house_interior.obj, mtl: /storage/emulated/0/jni_test/luxury_house_interior.mtl D/MarusyaObjLoader: End parsing -> /storage/emulated/0/jni_test/luxury_house_interior.obj D/MarusyaObjLoader: shapes size -> 116 D/MarusyaObjLoader: materials size -> 25 D/MarusyaObjLoader: vertices size -> 390021 D/MarusyaObjLoader: normals size -> 392763 D/MarusyaObjLoader: texcoords size -> 130796 D/MarusyaObjLoader: colors size -> 390021 D/MarusyaObjLoader: indices size -> 773121 D/MarusyaObjLoader: Time to parse -> 1364 ms D/MarusyaObjLoader: *********************************************************************************** 

void disableLog();

For the performance tests were read 50 3-D models of .obj format. It was done using pure Java and JNI. For pure Java was used this cool library https://github.com/javagl/Obj. The table below shows gain in speed for devices with different power when reading 3-D models of .obj format using pure Java and JNI.

Device	File size versus read time
Google Pixel 2 Qualcom Snapdragon 835 (4x2.35 GGz + 4x1.9 GGz), RAM 4 GB
Huawei HRY - LX1 HiSiliconKirin 710 (4x2.2 GGz + 4x1.7 GGz), RAM 3 GB)
LGE Nexus 5X Qualcom Snapdragon 808 (2x1.44 GGz + 2x1.82 GGz), RAM 2 GB)
Samsung SM - G930F Samsung Exynos 9 Octa 8890 (8x2.3 GGz), RAM 4 GB)
Samsung SM - N950F Samsung Exynos 9 Octa 8895 (4x2.3 GGz + 4x1.7 ГГц), RAM 6 GB)
Samsung SM - T820 Qualcomm Snapdragon 820 (2x2.15 GGz + 2x1.6 GGz), RAM 4 GB)
Xiaomi Mi A1 Qualcomm Snapdragon 625 (4x2.0 GGz), RAM 4 GB)
Xiaomi Redmi Note 5 Qualcom Snapdragon 636 (8x1.8 GGz), RAM 3 GB)

On average, the download speed using JNI is 4.71 times higher.

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.