AndreasLH's picture
upload repo
56bd2b5
import pickle
import matplotlib.pyplot as plt
import numpy as np
import torch
from matplotlib import pyplot as plt
from PIL import Image
import os
from cubercnn import util, vis
from detectron2.data.catalog import MetadataCatalog
from detectron2.data.detection_utils import convert_image_to_rgb
from detectron2.layers.nms import batched_nms
from detectron2.utils.visualizer import Visualizer
from cubercnn.data.generate_depth_maps import setup_depth_model, depth_of_images
def make_random_boxes(n_boxes=10):
# rotation_matrix = torch.rand(3,3)*2*torch.pi
rotation_matrix = torch.eye(3) # no rotation
# need xyz, whl, and pose (R)
# whl = torch.rand(3)*0.5
whl = torch.tensor([0.3, 0.3, 0.3])
xyz = torch.tensor([-0.1, 0, 1.7])
# xyz = torch.rand(3)*1
return xyz, whl, rotation_matrix
def proposals_3d_from_2d(image, pred2d):
with open('3dboxes/proposals/network_out.pkl', 'rb') as f:
batched_inputs, images, features, proposals, Ks, gt_instances, im_scales_ratio, instances = pickle.load(f)
n_boxes = 1
pred_xyz, pred_whl, pred_pose = make_random_boxes(n_boxes=n_boxes)
pred_xyzwhl = torch.cat((pred_xyz, pred_whl), dim=0)
pred_colors = torch.tensor([util.get_color(i) for i in range(n_boxes)])/255.0
pred_meshes = util.mesh_cuboid(pred_xyzwhl, pred_pose, pred_colors)
input_format = 'BGR'
img = batched_inputs[0]['image']
img = convert_image_to_rgb(img.permute(1, 2, 0), input_format)
img_3DPR = np.ascontiguousarray(img.copy()[:, :, [2, 1, 1]]) # BGR
input = batched_inputs[0]
K = torch.tensor(input['K'])
scale = input['height']/img.shape[0]
K_scaled = torch.tensor(
[[1/scale, 0 , 0], [0, 1/scale, 0], [0, 0, 1.0]],
dtype=torch.float32) @ K
# convert to lists
pred_meshes = [pred_meshes.__getitem__(i).detach() for i in range(len(pred_meshes))]
# horizontal stack 3D GT and pred left/right
# 2 box
box_size = min(len(proposals[0].proposal_boxes), 2)
v_pred = Visualizer(img, None)
v_pred = v_pred.overlay_instances(
boxes=proposals[0].proposal_boxes[0:box_size].tensor.cpu().numpy()
)
prop_img = v_pred.get_image()
img_3DPR = vis.draw_scene_view(prop_img, K_scaled.cpu().numpy(), pred_meshes, text=['3d box'], mode='front', blend_weight=0.0, blend_weight_overlay=0.85)
# vis_img_3d = img_3DPR[:, :, [2, 1, 0]] # RGB
vis_img_3d = img_3DPR.astype(np.uint8)
fig, ax = plt.subplots(); ax.imshow(vis_img_3d); ax.axis('off')
plt.savefig(f'3dboxes/proposals/figs/pred.png', bbox_inches='tight', dpi=300)
# visualize(batched_inputs, proposals, instances)
return
def visualize(batched_inputs, proposals, instances):
# taken from the class ROIHeads3D
"""
A function used to visualize images and proposals. It shows ground truth
bounding boxes on the original image and up to 20 top-scoring predicted
object proposals on the original image. Users can implement different
visualization functions for different models.
Args:
batched_inputs (list): a list that contains input to the model.
proposals (list): a list that contains predicted proposals. Both
batched_inputs and proposals should have the same length.
instances (list): a list that contains predicted RoIhead instances. Both
batched_inputs and proposals should have the same length.
"""
max_vis_prop = 2
device = 'cpu'
input_format = 'BGR'
# thing_classes = MetadataCatalog.get('omni3d_model').thing_classes
thing_classes = ['pedestrian', 'car', 'cyclist', 'van', 'truck', 'traffic cone', 'barrier', 'motorcycle', 'bicycle', 'bus', 'trailer', 'books', 'bottle', 'camera', 'cereal box', 'chair', 'cup', 'laptop', 'shoes', 'towel', 'blinds', 'window', 'lamp', 'shelves', 'mirror', 'sink', 'cabinet', 'bathtub', 'door', 'toilet', 'desk', 'box', 'bookcase', 'picture', 'table', 'counter', 'bed', 'night stand', 'pillow', 'sofa', 'television', 'floor mat', 'curtain', 'clothes', 'stationery', 'refrigerator', 'bin', 'stove', 'oven', 'machine']
num_classes = len(thing_classes)
for i, (input, prop, instances_i) in enumerate(zip(batched_inputs, proposals, instances)):
img = input["image"]
img = convert_image_to_rgb(img.permute(1, 2, 0), input_format)
img_3DGT = np.ascontiguousarray(img.copy()[:, :, [2, 1, 1]]) # BGR
img_3DPR = np.ascontiguousarray(img.copy()[:, :, [2, 1, 1]]) # BGR
'''
Visualize the 2D GT and proposal predictions
'''
v_gt = Visualizer(img, None)
v_gt = v_gt.overlay_instances(boxes=input["instances"].gt_boxes)
anno_img = v_gt.get_image()
box_size = min(len(prop.proposal_boxes), max_vis_prop)
v_pred = Visualizer(img, None)
v_pred = v_pred.overlay_instances(
boxes=prop.proposal_boxes[0:box_size].tensor.cpu().numpy()
)
prop_img = v_pred.get_image()
vis_img_rpn = np.concatenate((anno_img, prop_img), axis=1)
# fig, ax = plt.subplots(); ax.imshow(vis_img_rpn); ax.axis('off')
# plt.savefig(f'3dboxes/proposals/figs/vis_img_rpn_{i}.png', bbox_inches='tight', dpi=300)
'''
Visualize the 3D GT and predictions
'''
K = torch.tensor(input['K'], device=device)
scale = input['height']/img.shape[0]
fx, sx = (val.item()/scale for val in K[0, [0, 2]])
fy, sy = (val.item()/scale for val in K[1, [1, 2]])
K_scaled = torch.tensor(
[[1/scale, 0 , 0], [0, 1/scale, 0], [0, 0, 1.0]],
dtype=torch.float32, device=device
) @ K
gts_per_image = input["instances"]
gt_classes = gts_per_image.gt_classes
# Filter out irrelevant groundtruth
fg_selection_mask = (gt_classes != -1) & (gt_classes < num_classes)
gt_classes = gt_classes[fg_selection_mask]
gt_class_names = [thing_classes[cls_idx] for cls_idx in gt_classes]
gt_boxes = gts_per_image.gt_boxes.tensor[fg_selection_mask] # 2D boxes
gt_poses = gts_per_image.gt_poses[fg_selection_mask] # GT poses
# projected 2D center, depth, w, h, l, 3D center
gt_boxes3D = gts_per_image.gt_boxes3D[fg_selection_mask]
# this box may have been mirrored and scaled so
# we need to recompute XYZ in 3D by backprojecting.
gt_z = gt_boxes3D[:, 2]
gt_x3D = gt_z * (gt_boxes3D[:, 0] - sx)/fx
gt_y3D = gt_z * (gt_boxes3D[:, 1] - sy)/fy
# put together the GT boxes
gt_center_3D = torch.stack((gt_x3D, gt_y3D, gt_z)).T
gt_boxes3D_XYZ_WHL = torch.cat((gt_center_3D, gt_boxes3D[:, 3:6]), dim=1)
gt_colors = torch.tensor(
[util.get_color(i) for i in range(len(gt_boxes3D_XYZ_WHL))],
device=device
)/255.0
gt_meshes = util.mesh_cuboid(gt_boxes3D_XYZ_WHL, gt_poses, gt_colors)
# perform a simple NMS, which is not cls dependent.
keep = batched_nms(
instances_i.pred_boxes.tensor,
instances_i.scores,
torch.zeros(len(instances_i.scores), dtype=torch.long, device=instances_i.scores.device),
0.5 # this should come from roi_heads.nms_thresh
)
keep = keep[:max_vis_prop]
num_to_visualize = len(keep)
pred_xyzwhl = torch.cat((instances_i.pred_center_cam[keep], instances_i.pred_dimensions[keep]), dim=1)
pred_pose = instances_i.pred_pose[keep]
pred_colors = torch.tensor(
[util.get_color(i) for i in range(num_to_visualize)],
device=device
)/255.0
pred_boxes = instances_i.pred_boxes[keep]
pred_scores = instances_i.scores[keep]
pred_classes = instances_i.pred_classes[keep]
pred_class_names = ['{} {:.2f}'.format(thing_classes[cls_idx], score) for cls_idx, score in zip(pred_classes, pred_scores)]
pred_meshes = util.mesh_cuboid(pred_xyzwhl, pred_pose, pred_colors)
# print(pred_xyzwhl)
# convert to lists
pred_meshes = [pred_meshes.__getitem__(i).detach() for i in range(len(pred_meshes))]
gt_meshes = [gt_meshes.__getitem__(i) for i in range(len(gt_meshes))]
img_3DPR = vis.draw_scene_view(anno_img, K_scaled.cpu().numpy(), pred_meshes, text=pred_class_names, mode='front', blend_weight=0.0, blend_weight_overlay=0.85)
img_3DGT = vis.draw_scene_view(img_3DGT, K_scaled.cpu().numpy(), gt_meshes, text=gt_class_names, mode='front', blend_weight=0.0, blend_weight_overlay=0.85)
# horizontal stack 3D GT and pred left/right
img_3DGT = img_3DGT[:, :, [2, 1, 0]] # RGB
vis_img_3d = np.concatenate((img_3DGT, img_3DPR), axis=1)
vis_img_3d = vis_img_3d.astype(np.uint8)
fig, ax = plt.subplots(); ax.imshow(vis_img_3d); ax.axis('off')
plt.savefig(f'3dboxes/proposals/figs/vis_img_3d_{i}.png', bbox_inches='tight', dpi=300)
if __name__ == "__main__":
# proposals_3d_from_2d(None, None)
with open('ProposalNetwork/proposals/network_out.pkl', 'rb') as f:
batched_inputs, images, features, proposals, Ks, gt_instances, im_scales_ratio, instances = pickle.load(f)
n_boxes = 1
pred_xyz, pred_whl, pred_pose = make_random_boxes(n_boxes=n_boxes)
pred_xyzwhl = torch.cat((pred_xyz, pred_whl), dim=0)
pred_colors = torch.tensor([util.get_color(i) for i in range(n_boxes)])/255.0
pred_meshes = util.mesh_cuboid(pred_xyzwhl, pred_pose, pred_colors)
input_format = 'BGR'
img = batched_inputs[0]['image']
img = convert_image_to_rgb(img.permute(1, 2, 0), input_format)
depth_model = 'zoedepth'
# the local:: thing of the model path is just to indicate that the model is loaded local storage
pretrained_resource = 'local::depth/checkpoints/depth_anything_metric_depth_indoor.pt'
model = setup_depth_model(depth_model, pretrained_resource)
resized_pred = depth_of_images(img, model)
plt.matshow(resized_pred)
plt.savefig(os.path.join('/work3/s194369/3dod/3dboxes/output/trash', 'depth_img.png'),dpi=300, bbox_inches='tight')
plt.show()