Coded XML interpreter. Just need to extract animation pose matrices and turn them into quaternions and positions for keyframe interpolation.
Converting matrix to quaternion using GLM:
glm::quat new_quat = glm::toQuat(bone_mat2);
For simplicity, multiple COLLADA model files can be used for different animations. Still using JSON files for vertices, normals and UV information, but ResMan should at some point be able to gather everything solely from one COLLADA file.