Merge pull request #69 from kaseris/notebooks/amass

AMASS dataset notebook presentation and comments

notebooks/amass_visualization.ipynb

@@ -0,0 +1,244 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# Experimenting with the AMASS dataset"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "%load_ext autoreload\n",
+ "%autoreload 2\n",
+ "%matplotlib notebook\n",
+ "%matplotlib inline"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import os.path as osp\n",
+ "\n",
+ "import torch\n",
+ "import numpy as np\n",
+ "\n",
+ "from human_body_prior.tools.omni_tools import copy2cpu as c2c\n",
+ "\n",
+ "support_dir = '/home/kaseris/Documents/amass/support_data'\n",
+ "comp_device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "amass_npz_fname = osp.join(support_dir, 'github_data/dmpl_sample.npz')\n",
+ "bdata = np.load(amass_npz_fname)\n",
+ "\n",
+ "subject_gender = bdata['gender'].tostring().decode('utf-8')"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "Let's see what's inside the bdata\n",
+ "\n",
+ "- What are the `dmpls`? Some body parameters????\n",
+ "- What are the `betas`? Body parameters"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "print(f'The available keys from the read file are: {list(bdata.keys())}')\n",
+ "print(f'betas: {bdata[\"betas\"].shape}')\n",
+ "print(f'dmpls: {bdata[\"dmpls\"].shape}')\n",
+ "print(f'The subject gender is: {subject_gender}')"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "from human_body_prior.body_model.body_model import BodyModel\n",
+ "\n",
+ "dmpl_fname = osp.join(support_dir, f'body_models/dmpls/{subject_gender}/model.npz')\n",
+ "bm_fname = osp.join(support_dir, f'body_models/smplh/{subject_gender}/model.npz')\n",
+ "\n",
+ "num_betas = 16 # number of body parameters\n",
+ "num_dmpls = 8 # number of DMPL parameters\n",
+ "\n",
+ "bm = BodyModel(bm_fname=bm_fname, num_betas=num_betas, num_dmpls=num_dmpls, dmpl_fname=dmpl_fname).to(comp_device)\n",
+ "faces = c2c(bm.f)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "The provided sample data has the original mocap marker data.\n",
+ "\n",
+ "In the real AMASS dataset, we include only markers for the test set.\n",
+ "\n",
+ "For the rest of the subsets you can obtain the marker data from their respective websites.\n",
+ "\n",
+ "In the following we make PyTorch tensors for parameters controlling different part of the body model."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "time_length = len(bdata['trans'])\n",
+ "body_parms = {\n",
+ " 'root_orient': torch.Tensor(bdata['poses'][:, :3]).to(comp_device), # controls the global root orientation\n",
+ " 'pose_body': torch.Tensor(bdata['poses'][:, 3:66]).to(comp_device), # controls the body\n",
+ " 'pose_hand': torch.Tensor(bdata['poses'][:, 66:]).to(comp_device), # controls the finger articulation\n",
+ " 'trans': torch.Tensor(bdata['trans']).to(comp_device), # controls the global body position\n",
+ " 'betas': torch.Tensor(np.repeat(bdata['betas'][:num_betas][np.newaxis], repeats=time_length, axis=0)).to(comp_device), # controls the body shape. Body shape is static\n",
+ " 'dmpls': torch.Tensor(bdata['dmpls'][:, :num_dmpls]).to(comp_device) # controls soft tissue dynamics\n",
+ "}\n",
+ "print('Body parameter vector shapes: \\n{}'.format(' \\n'.join(['{}: {}'.format(k,v.shape) for k,v in body_parms.items()])))\n",
+ "print('time_length = {}'.format(time_length))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import trimesh\n",
+ "from body_visualizer.tools.vis_tools import colors\n",
+ "from body_visualizer.mesh.mesh_viewer import MeshViewer\n",
+ "from body_visualizer.mesh.sphere import points_to_spheres\n",
+ "from body_visualizer.tools.vis_tools import show_image\n",
+ "\n",
+ "imw, imh=1600, 1600\n",
+ "mv = MeshViewer(width=imw, height=imh, use_offscreen=True)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "body_pose_beta = bm(**{k:v for k,v in body_parms.items() if k in ['pose_body', 'betas']})\n",
+ "\n",
+ "def vis_body_pose_beta(fId = 0):\n",
+ " body_mesh = trimesh.Trimesh(vertices=c2c(body_pose_beta.v[fId]), faces=faces, vertex_colors=np.tile(colors['grey'], (6890, 1)))\n",
+ " mv.set_static_meshes([body_mesh])\n",
+ " body_image = mv.render(render_wireframe=False)\n",
+ " show_image(body_image)\n",
+ "\n",
+ "vis_body_pose_beta(fId=0)\n",
+ "\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "Now the model has a more realistic hand posture"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "body_pose_hand = bm(**{k:v for k,v in body_parms.items() if k in ['pose_body', 'betas', 'pose_hand']})\n",
+ "\n",
+ "def vis_body_pose_hand(fId = 0):\n",
+ " body_mesh = trimesh.Trimesh(vertices=c2c(body_pose_hand.v[fId]), faces=faces, vertex_colors=np.tile(colors['grey'], (6890, 1)))\n",
+ " mv.set_static_meshes([body_mesh])\n",
+ " body_image = mv.render(render_wireframe=False)\n",
+ " show_image(body_image)\n",
+ "\n",
+ "vis_body_pose_hand(fId=0)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "Visualize the body joints"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def vis_body_joints(fId = 0):\n",
+ " joints = c2c(body_pose_hand.Jtr[fId])\n",
+ " joints_mesh = points_to_spheres(joints, point_color = colors['red'], radius=0.005)\n",
+ "\n",
+ " mv.set_static_meshes([joints_mesh])\n",
+ " body_image = mv.render(render_wireframe=False)\n",
+ " show_image(body_image)\n",
+ "\n",
+ "vis_body_joints(fId=0)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "body_dmpls = bm(**{k:v for k,v in body_parms.items() if k in ['pose_body', 'betas', 'pose_hand', 'dmpls']})\n",
+ "\n",
+ "def vis_body_dmpls(fId = 0):\n",
+ " body_mesh = trimesh.Trimesh(vertices=c2c(body_dmpls.v[fId]), faces=faces, vertex_colors=np.tile(colors['grey'], (6890, 1)))\n",
+ " mv.set_static_meshes([body_mesh])\n",
+ " body_image = mv.render(render_wireframe=False)\n",
+ " show_image(body_image)\n",
+ "\n",
+ "vis_body_dmpls(fId=0)"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "scraping",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.10.13"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

requirements.txt

@@ -7,4 +7,8 @@ matplotlib
 randomname
 pyyaml
 tensorboard
-discord
+discord
+git+https://github.com/nghorbani/configer
+git+https://github.com/nghorbani/human_body_prior
+git+https://github.com/nghorbani/amass
+git+https://github.com/nghorbani/body_visualizer