def run(**kwargs):
print('a')
#get configuration
try:
config = importlib.import_module('config.'+kwargs['config']);
opt = config.__dict__;
for k in kwargs.keys():
if not kwargs[k] is None:
opt[k] = kwargs[k];
except Exception as e:
print(e);
traceback.print_exc();
exit();
#get network
try:
m = importlib.import_module('net.'+opt['net']);
net = m.Net(**opt);
if torch.cuda.is_available():
net = net.cuda();
except Exception as e:
print(e);
traceback.print_exc();
exit();
#get dataset
try:
m = importlib.import_module('util.dataset.'+opt['dataset']);
train_data = m.Data(opt,True);
val_data = m.Data(opt,False);
train_load = DataLoader(train_data,batch_size=opt['batch_size'],shuffle=True,num_workers=opt['workers']);
val_load = DataLoader(val_data,batch_size=opt['batch_size'],shuffle=False,num_workers=opt['workers']);
except Exception as e:
print(e);
traceback.print_exc();
exit();
if opt['model']!='':
partial_restore(net,opt['model']);
print("Previous weights loaded");
if 'train' in opt['user_key']:
load = train_load;
else:
load = val_load;
if opt['model']!='':
outdir = os.path.dirname(opt['model'])+os.sep+'view_'+opt['user_key'];
if not os.path.exists(outdir):
os.mkdir(outdir);
for i, data in enumerate(load,0):
print(i,'/',len(train_data)//opt['batch_size']);
data2cuda(data);
net.eval();
with torch.no_grad():
out = net(data);
img = data[0].data.cpu().numpy();
box2d_src = data[1].data.cpu().numpy();
box3d_src = data[2].data.cpu().numpy();
box2d_tgt = data[3].data.cpu().numpy();
box3d_tgt = data[4].data.cpu().numpy();
r = data[5].data.cpu().numpy();
gts = data[6].data.cpu().numpy();
y = out['y'].data.cpu().numpy();
tri = box_face;
num = box3d_src.shape[0];
col = 4;
row = num // col;
for ri in range(row):
for cj in range(col):
ni = ri*col+cj;
fig = plt.figure(figsize=(48,16));
#
ax = fig.add_subplot(132,projection='3d');
ax.view_init(elev=20, azim=0)
ax.plot_trisurf(box3d_tgt[ni,...,0],box3d_tgt[ni,...,1],tri,box3d_tgt[ni,...,2],color=(0,0,1,0.1));
ax.plot_trisurf(box3d_src[ni,...,0],box3d_src[ni,...,1],tri,box3d_src[ni,...,2],color=(0,1,0,0.1));
ygt = gts[ni,...];
ygt *= np.pi;
ygt[1] *= 2;
c3d1 = recon(box3d_src[ni,...],r[ni,...],ygt);
ax.plot(c3d1[:,0],c3d1[:,1],c3d1[:,2],color='k');
ymap = y[ni,...];
ymap *= np.pi;
ymap[1] *= 2;
c3d2 = recon(box3d_src[ni,...],r[ni,...],ymap);
ax.plot(c3d2[:,0],c3d2[:,1],c3d2[:,2],color='r');
ax.scatter(box3d_src[ni,0:4,0],box3d_src[ni,0:4,1],box3d_src[ni,0:4,2],color='b',marker='*');
ax.scatter(box3d_src[ni,4:8,0],box3d_src[ni,4:8,1],box3d_src[ni,4:8,2],color='c',marker='*');
ax.scatter(box3d_tgt[ni,0:4,0],box3d_tgt[ni,0:4,1],box3d_tgt[ni,0:4,2],color='k',marker='x');
ax.scatter(box3d_tgt[ni,4:8,0],box3d_tgt[ni,4:8,1],box3d_tgt[ni,4:8,2],color='r',marker='x');
ax.set_aspect('equal', adjustable='box');
#
ax = fig.add_subplot(133,projection='3d');
ax.view_init(elev=20, azim=90)
ax.plot_trisurf(box3d_tgt[ni,...,0],box3d_tgt[ni,...,1],tri,box3d_tgt[ni,...,2],color=(0,0,1,0.1));
ax.plot_trisurf(box3d_src[ni,...,0],box3d_src[ni,...,1],tri,box3d_src[ni,...,2],color=(0,1,0,0.1));
ax.plot(c3d1[:,0],c3d1[:,1],c3d1[:,2],color='k');
ax.plot(c3d2[:,0],c3d2[:,1],c3d2[:,2],color='r');
ax.scatter(box3d_src[ni,0:4,0],box3d_src[ni,0:4,1],box3d_src[ni,0:4,2],color='b',marker='*');
ax.scatter(box3d_src[ni,4:8,0],box3d_src[ni,4:8,1],box3d_src[ni,4:8,2],color='c',marker='*');
ax.scatter(box3d_tgt[ni,0:4,0],box3d_tgt[ni,0:4,1],box3d_tgt[ni,0:4,2],color='k',marker='x');
ax.scatter(box3d_tgt[ni,4:8,0],box3d_tgt[ni,4:8,1],box3d_tgt[ni,4:8,2],color='r',marker='x');
ax.set_aspect('equal', adjustable='box');
#
ax = fig.add_subplot(131);
ax.set_aspect('equal', adjustable='box');
ax.imshow(img[ni,...]);
ax.scatter(box2d_src[ni,0:4,0],box2d_src[ni,0:4,1],color='b',marker='*');
ax.scatter(box2d_src[ni,4:8,0],box2d_src[ni,4:8,1],color='c',marker='*');
ax.scatter(box2d_tgt[ni,0:4,0],box2d_tgt[ni,0:4,1],color='k',marker='x');
ax.scatter(box2d_tgt[ni,4:8,0],box2d_tgt[ni,4:8,1],color='r',marker='x');
ax.set_aspect('equal', adjustable='box');
c2d1 = proj(mv(c3d1));
c2d2 = proj(mv(c3d2));
ax.plot(c2d1[:,0],c2d1[:,1],color='k');
ax.plot(c2d2[:,0],c2d2[:,1],color='r');
if opt['model']!='':
plt.savefig(os.path.join(outdir,"_%04d_%04d.png"%(i,ni)));
if opt['ply']:
plt.show();
plt.close(fig);
#run the code
'''
def run(**kwargs):
global iternum
#get configuration
try:
config = importlib.import_module('config.' + kwargs['config'])
opt = config.__dict__
for k in kwargs.keys():
if not kwargs[k] is None:
opt[k] = kwargs[k]
except Exception as e:
print(e)
traceback.print_exc()
exit()
iternum = opt['nepoch']
#get network
try:
m = importlib.import_module('net.' + opt['net'])
net = m.Net(**opt)
if torch.cuda.is_available():
net = net.cuda()
except Exception as e:
print(e)
traceback.print_exc()
exit()
#get dataset
try:
m = importlib.import_module('util.dataset.' + opt['dataset'])
train_data = m.Data(opt, 'train')
val_data = m.Data(opt, 'test')
train_load = DataLoader(train_data,
batch_size=opt['batch_size'],
shuffle=True,
num_workers=opt['workers'])
val_load = DataLoader(val_data,
batch_size=opt['batch_size'],
shuffle=False,
num_workers=opt['workers'])
except Exception as e:
print(e)
traceback.print_exc()
exit()
print('bs:', opt['batch_size'])
if opt['model'] != '':
partial_restore(net, opt['model'])
print("Previous weights loaded")
if opt['model'] != '':
outdir = os.path.dirname(opt['model']) + os.sep + 'view'
if not os.path.exists(outdir):
os.mkdir(outdir)
#run the code
optim = eval('optim.' + opt['optim'])(config.parameters(net),
lr=opt['lr'],
weight_decay=opt['weight_decay'])
tri = box_face
fidx = tri
T = np.dtype([("n", np.uint8), ("i0", np.int32), ('i1', np.int32),
('i2', np.int32)])
face = np.zeros(shape=[2 * fidx.shape[0]], dtype=T)
for i in range(2 * fidx.shape[0]):
if i < fidx.shape[0]:
face[i] = (3, fidx[i, 0], fidx[i, 1], fidx[i, 2])
else:
face[i] = (3, fidx[i - fidx.shape[0], 0] + 8,
fidx[i - fidx.shape[0], 1] + 8,
fidx[i - fidx.shape[0], 2] + 8)
for i, data in enumerate(val_load, 0):
print(i, '/', len(val_data))
data2cuda(data)
img = data[0].data.cpu().numpy()
msks = data[1].data.cpu().numpy()
mskt = data[2].data.cpu().numpy()
ygt = data[3].data.cpu().numpy()
vgt = data[4].data.cpu().numpy()
sgt = data[-3].data.cpu().numpy()
net.eval()
with torch.no_grad():
out = net(data)
acc = config.accuracy(data, out)
id = data[-2]
tb = out['tb'].data.cpu().numpy()
sb = out['sb'].data.cpu().numpy()
#y = out['y'].data.cpu().numpy();
cat = data[-1]
err = acc['t'].data.cpu().numpy()
#
for tagi, tag in enumerate(id):
cpath = os.path.join(
outdir, '_' + cat[tagi] + '_%04d' % i + '_%02d' % tagi)
if not os.path.exists(cpath):
os.mkdir(cpath)
im = Image.fromarray((img[tagi, ...] * 255).astype(np.uint8),
mode='RGB')
im.save(os.path.join(cpath, '_000_img.png'))
if opt['mode'] == 'full':
im.save(os.path.join(cpath, '_000_input.png'))
elif opt['mode'] == 'part':
iim = img[tagi, ...] * (
(msks[tagi, ...] + mskt[tagi, ...]).reshape(224, 224, 1))
iim = Image.fromarray((iim * 255).astype(np.uint8), mode='RGB')
iim.save(os.path.join(cpath, '_000_input.png'))
mks = Image.fromarray((msks[tagi, ...] * 255).astype(np.uint8),
mode='L')
mks.save(os.path.join(cpath, '_000_msks.png'))
mkt = Image.fromarray((mskt[tagi, ...] * 255).astype(np.uint8),
mode='L')
mkt.save(os.path.join(cpath, '_000_mskt.png'))
ptsa, ptsb = parse(vgt[tagi, ...])
center = vgt[tagi, 12:15]
print('L2:%f' % (err[tagi, ...]),
file=open(os.path.join(cpath, '_000_log.txt'), 'w'))
print('id:%s' % (id[tagi]),
file=open(os.path.join(cpath, '_000_meta.txt'), 'w'))
st = sgt[tagi, :]
st = st[np.newaxis, :]
ptgt = np.concatenate([ptsa, ptsb], axis=0)
ptgt += st
write_ply(os.path.join(cpath, '_000_gt.ply'),
points=pd.DataFrame(ptgt),
faces=pd.DataFrame(face))
ptout = np.concatenate([sb[tagi, ...], tb[tagi, ...]], axis=0)
ptout += st
write_ply(os.path.join(cpath, '_000_out.ply'),
points=pd.DataFrame(ptout),
faces=pd.DataFrame(face))
def run(**kwargs):
global iternum
#get configuration
try:
config = importlib.import_module('config.' + kwargs['config'])
opt = config.__dict__
for k in kwargs.keys():
if not kwargs[k] is None:
opt[k] = kwargs[k]
except Exception as e:
print(e)
traceback.print_exc()
exit()
iternum = opt['nepoch']
#get network
try:
m = importlib.import_module('net.' + opt['net'])
net = m.Net(**opt)
if torch.cuda.is_available():
net = net.cuda()
except Exception as e:
print(e)
traceback.print_exc()
exit()
#get dataset
try:
m = importlib.import_module('util.dataset.' + opt['dataset'])
train_data = m.Data(opt, True)
val_data = m.Data(opt, False)
train_load = DataLoader(train_data,
batch_size=opt['batch_size'],
shuffle=True,
num_workers=opt['workers'])
val_load = DataLoader(val_data,
batch_size=opt['batch_size'],
shuffle=False,
num_workers=opt['workers'])
except Exception as e:
print(e)
traceback.print_exc()
exit()
if opt['model'] != '':
partial_restore(net, opt['model'])
print("Previous weights loaded")
if 'train' in opt['user_key']:
load = train_load
print(train_data.datapath[0])
else:
load = val_load
print(val_data.datapath[0])
if opt['model'] != '':
outdir = os.path.dirname(
opt['model']) + os.sep + 'view_' + opt['user_key']
if not os.path.exists(outdir):
os.mkdir(outdir)
for i, data in enumerate(load, 0):
data2cuda(data)
d = data
#
net.eval()
with torch.no_grad():
out = net(data)
#
img = data[0].data.cpu().numpy()
box_src = data[1].data.cpu().numpy()
box_tgt = data[2].data.cpu().numpy()
center = data[3].data.cpu().numpy()
r = data[4].data.cpu().numpy()
gt = data[5].data.cpu().numpy()
yout = out['y'].data.cpu().numpy()
map = out['dmap'].permute(0, 2, 3, 1).data.cpu().numpy()
#run the code
#optim = eval('optim.'+opt['optim'])(config.parameters(net),lr=opt['lr'],weight_decay=opt['weight_decay']);
tri = box_face
global pv
num = box_src.shape[0]
col = int(np.sqrt(num))
row = num // col
for ri in range(row):
for cj in range(col):
ni = ri * col + cj
fig = plt.figure(figsize=(48, 16))
#===========================================
y = gt[ni, ...].copy()
y *= np.pi
y[1] *= 2
c3d = recon(center[ni, ...], r[ni, ...], y)
#===========================================
ymap = yout[ni, ...]
ymap *= np.pi
ymap[1] *= 2
c3do = recon(center[ni, ...], r[ni, ...], ymap)
c2do = proj(mv(c3do))
#===========================================
ax = fig.add_subplot(1, 5, 1)
ax.imshow(img[ni, ...])
c2d = proj(mv(c3d))
line = ax.plot(c2do[:, 0], c2do[:, 1], color='r')
pv.extend(line)
ax.plot(c2d[:, 0], c2d[:, 1], color='k')
#===========================================
ax = fig.add_subplot(1, 5, 2)
im1 = norm_im(map[ni, :, :, :3])
im1 = ax.imshow(im1)
pv.append(im1)
#===========================================
ax = fig.add_subplot(1, 5, 3)
im2 = norm_im(map[ni, :, :, 3:])
im2 = ax.imshow(im2)
pv.append(im2)
#===========================================
ax = fig.add_subplot(1, 5, 4, projection='3d')
ax.view_init(elev=20, azim=90)
ax.set_aspect('equal', adjustable='box')
ax.plot_trisurf(box_tgt[ni, ..., 0],
box_tgt[ni, ..., 1],
tri,
box_tgt[ni, ..., 2],
color=(0, 0, 1, 0.1))
ax.plot_trisurf(box_src[ni, ..., 0],
box_src[ni, ..., 1],
tri,
box_src[ni, ..., 2],
color=(0, 1, 0, 0.1))
line = ax.plot(c3do[:, 0], c3do[:, 1], c3do[:, 2], color='r')
#==========================================
pv.extend(line)
ax.plot(c3d[:, 0], c3d[:, 1], c3d[:, 2], color='k')
#==========================================
ax = fig.add_subplot(1, 5, 5, projection='3d')
ax.set_aspect('equal', adjustable='box')
ax.plot_trisurf(box_tgt[ni, ..., 0],
box_tgt[ni, ..., 1],
tri,
box_tgt[ni, ..., 2],
color=(0, 0, 1, 0.1))
ax.plot_trisurf(box_src[ni, ..., 0],
box_src[ni, ..., 1],
tri,
box_src[ni, ..., 2],
color=(0, 1, 0, 0.1))
line = ax.plot(c3do[:, 0], c3do[:, 1], c3do[:, 2], color='r')
#===========================================
pv.extend(line)
ax.plot(c3d[:, 0], c3d[:, 1], c3d[:, 2], color='k')
#===========================================
if opt['model'] != '':
plt.savefig(
os.path.join(outdir, "_%04d_%04d.png" % (i, ni)))
if opt['ply']:
plt.show()
plt.close(fig)
#============
#run the code
'''
def run(**kwargs):
#get configuration
try:
config = importlib.import_module('config.'+kwargs['config']);
opt = config.__dict__;
for k in kwargs.keys():
if not kwargs[k] is None:
opt[k] = kwargs[k];
except Exception as e:
print(e);
traceback.print_exc();
exit();
#get network
try:
m = importlib.import_module('net.'+opt['net']);
net = m.Net(**opt);
if torch.cuda.is_available():
net = net.cuda();
except Exception as e:
print(e);
traceback.print_exc();
exit();
#get dataset
try:
m = importlib.import_module('util.dataset.'+opt['dataset']);
train_data = m.Data(opt,opt['user_key']);
train_load = DataLoader(train_data,batch_size=opt['batch_size'],shuffle=False,num_workers=opt['workers']);
except Exception as e:
print(e);
traceback.print_exc();
exit();
if opt['model']!='':
partial_restore(net,opt['model']);
print("Previous weights loaded");
if opt['model']!='':
outdir = os.path.dirname(opt['model'])+os.sep+'view';
if not os.path.exists(outdir):
os.mkdir(outdir);
#
root = os.path.join(opt['data_path'],'test');
cat_lst = os.listdir(root);
for c in cat_lst:
path = os.path.join(root,c);
cout = os.path.join(outdir,'_'+c);
if not os.path.exists(cout):
os.mkdir(cout);
if os.path.isdir(path):
f_lst = os.listdir(path);
cnt = 0;
for i,f in enumerate(f_lst):
if f.endswith('.h5'):
fopath = os.path.join(cout,'_%04d'%cnt);
h5f = h5py.File(os.path.join(path,f),'r');
if not os.path.exists(fopath):
os.mkdir(fopath);
img = np.array(h5f['img']);
Image.fromarray((img*255.0).astype(np.uint8),mode='RGB').save(os.path.join(fopath,'_input.png'));
msk = np.array(h5f['msk']);
touch = np.array(h5f['touch']);
box = np.array(h5f['box']);
write_box(box,os.path.join(fopath,'_gt.ply'));
obox = infer_box(net,img,msk,touch,box);
write_box(obox,os.path.join(fopath,'_out.ply'));
h5f.close();
cnt += 1;
if cnt > 10:
break;
def run(**kwargs):
#get configuration
try:
config = importlib.import_module('config.' + kwargs['config'])
opt = config.__dict__
for k in kwargs.keys():
if not kwargs[k] is None:
opt[k] = kwargs[k]
except Exception as e:
print(e)
traceback.print_exc()
exit()
#get network
try:
m = importlib.import_module('net.' + opt['net'])
net = m.Net(**opt)
if torch.cuda.is_available():
net = net.cuda()
except Exception as e:
print(e)
traceback.print_exc()
exit()
#get dataset
try:
m = importlib.import_module('util.dataset.' + opt['dataset'])
train_data = m.Data(opt, True)
train_load = DataLoader(train_data,
batch_size=opt['batch_size'],
shuffle=False,
num_workers=opt['workers'])
except Exception as e:
print(e)
traceback.print_exc()
exit()
if opt['model'] != '':
partial_restore(net, opt['model'])
print("Previous weights loaded")
bi = int(opt['user_key'])
if opt['model'] != '':
outdir = os.path.dirname(opt['model']) + os.sep + 'view_%d' % bi
if not os.path.exists(outdir):
os.mkdir(outdir)
#
input_size = opt['input_size']
idx = opt['part_idx']
tri = box_face
for i, data in enumerate(train_load, 0):
print(i, '/',
len(train_data) // opt['batch_size'])
data2cuda(data)
net.eval()
zs = np.linspace(-3, 3, 20).astype(np.float32)
xx = data[0][bi, idx].contiguous().view(1, -1)
ox = data[0][bi, :].contiguous().view(1, -1)
cat = data[1][bi]
print(cat)
#==================================
ptsa, ptsb = parse(ox.data.cpu().numpy()[0, :])
#==================================
fig = plt.figure(figsize=(9.6, 4.8))
ax = fig.add_subplot(1, 2, 1, projection='3d')
ax.view_init(elev=20, azim=90)
ax.set_aspect('equal', adjustable='box')
ax.set_xlim([-1, 1])
ax.set_ylim([-1, 1])
ax.set_zlim([-1, 1])
#
ax.plot_trisurf(ptsa[..., 0],
ptsa[..., 2],
tri,
ptsa[..., 1],
color=(0, 0, 1, 0.1))
ax.plot_trisurf(ptsb[..., 0],
ptsb[..., 2],
tri,
ptsb[..., 1],
color=(0, 1, 0, 0.1))
ax = fig.add_subplot(1, 2, 2, projection='3d')
ax.set_aspect('equal', adjustable='box')
ax.set_xlim([-1, 1])
ax.set_ylim([-1, 1])
ax.set_zlim([-1, 1])
#
ax.plot_trisurf(ptsa[..., 0],
ptsa[..., 2],
tri,
ptsa[..., 1],
color=(0, 0, 1, 0.1))
ax.plot_trisurf(ptsb[..., 0],
ptsb[..., 2],
tri,
ptsb[..., 1],
color=(0, 1, 0, 0.1))
plt.savefig(os.path.join(outdir, '_%d_%s_input.png' % (i, cat)))
plt.close(fig)
outdiri = os.path.join(outdir, '_%d_%s_output' % (i, cat))
if not os.path.exists(outdiri):
os.mkdir(outdiri)
for zi in range(opt['z_size']):
for ri in range(20):
fig = plt.figure(figsize=(9.6, 4.8))
z = np.zeros([1, opt['z_size']], dtype=np.float32)
z[0, zi] = zs[ri]
z = torch.from_numpy(z).cuda()
with torch.no_grad():
r = net.decode(xx, z)
x = r.data.cpu().numpy()[0, :]
ptsa, ptsb = parse(x)
#==================
ax = fig.add_subplot(1, 2, 1, projection='3d')
ax.view_init(elev=20, azim=90)
ax.set_aspect('equal', adjustable='box')
ax.set_xlim([-1, 1])
ax.set_ylim([-1, 1])
ax.set_zlim([-1, 1])
#
ax.plot_trisurf(ptsa[..., 0],
ptsa[..., 2],
tri,
ptsa[..., 1],
color=(0, 0, 1, 0.1))
ax.plot_trisurf(ptsb[..., 0],
ptsb[..., 2],
tri,
ptsb[..., 1],
color=(0, 1, 0, 0.1))
#
ax = fig.add_subplot(1, 2, 2, projection='3d')
ax.set_aspect('equal', adjustable='box')
ax.set_xlim([-1, 1])
ax.set_ylim([-1, 1])
ax.set_zlim([-1, 1])
#
ax.plot_trisurf(ptsa[..., 0],
ptsa[..., 2],
tri,
ptsa[..., 1],
color=(0, 0, 1, 0.1))
ax.plot_trisurf(ptsb[..., 0],
ptsb[..., 2],
tri,
ptsb[..., 1],
color=(0, 1, 0, 0.1))
plt.savefig(os.path.join(outdiri, '_%d_%f.png' % (zi, zs[ri])))
plt.close(fig)
def run(**kwargs):
global iternum
#get configuration
try:
config = importlib.import_module('config.' + kwargs['config'])
opt = config.__dict__
for k in kwargs.keys():
if not kwargs[k] is None:
opt[k] = kwargs[k]
except Exception as e:
print(e)
traceback.print_exc()
exit()
#get network
try:
m = importlib.import_module('net.' + opt['touch_net'])
touchnet = m.Net(**opt)
#
m = importlib.import_module('net.' + opt['box_net'])
boxnet = m.Net(**opt)
#
m = importlib.import_module('net.' + opt['touchpt_net'])
touchptnet = m.Net(**opt)
#
if torch.cuda.is_available():
touchnet = touchnet.cuda()
touchnet.eval()
boxnet = boxnet.cuda()
boxnet.eval()
touchptnet = touchptnet.cuda()
touchptnet.eval()
#
partial_restore(touchnet, opt['touch_model'])
partial_restore(boxnet, opt['box_model'])
partial_restore(touchptnet, opt['touchpt_model'])
except Exception as e:
print(e)
traceback.print_exc()
exit()
#get dataset
dpath = os.path.join(opt['data_path'], 'test')
opath = './log/joinN'
if not os.path.exists(opath):
os.makedirs(opath)
cat_lst = os.listdir(dpath)
#cat_lst = ['Chair','Table'];
eval_log = open('./log/joinN/' + opt['mode'] + '_eval.csv', 'w')
eval_all_sum = 0.0
eval_all_cnt = 0.0
for cat in cat_lst:
eval_cat_sum = 0.0
eval_cat_cnt = 0.0
cpath = os.path.join(opt['data_path'], 'test', cat)
copath = os.path.join(opath, 'joinN_' + cat + '_' + opt['mode'])
if not os.path.exists(copath):
os.mkdir(copath)
slst = os.listdir(cpath)
for f in slst:
id = os.path.basename(f).split('.')[-2]
fopath = os.path.join(copath, '_' + id)
if not os.path.exists(fopath):
os.mkdir(fopath)
logf = open(os.path.join(fopath, '_000_000_log.txt'), 'w')
h5f = h5py.File(os.path.join(cpath, f), 'r')
img = np.array(h5f['img'])
msk = np.array(h5f['msk'])
smsk = np.array(h5f['smsk'])
box = np.array(h5f['box'])
num = box.shape[0]
bdata = []
msk_lst = []
imgs_lst = []
imgt_lst = []
smsk_lst = []
tmsk_lst = []
box_lst = []
#for each part
rate = opt['user_rate']
for i in range(num):
msk_rate = (np.sum(msk[i, ...]) / np.sum(smsk[i, ...]))
if msk_rate > rate:
msk_lst.append(msk[i, ...])
imgs, mski = msk_center(img, msk[i, ...])
imgs_lst.append(imgs)
smsk_lst.append(mski)
imgt_lst.append(imgs)
tmsk_lst.append(mski)
box_lst.append(box[i, ...])
im = Image.fromarray((imgs_lst[-1] * 255).astype(np.uint8),
'RGB')
im.save(
os.path.join(fopath,
'_000_%03d_img.png' % (len(imgs_lst))))
imt = Image.fromarray(
(smsk_lst[-1] * 255).astype(np.uint8), 'L')
imt.save(
os.path.join(fopath,
'_001_%03d_msk.png' % (len(smsk_lst))))
#
imgs = np.stack(imgs_lst, axis=0)
smsk = np.stack(smsk_lst, axis=0)
imgt = np.stack(imgt_lst, axis=0)
tmsk = np.stack(tmsk_lst, axis=0)
bdata.append(torch.from_numpy(imgs).cuda())
bdata.append(torch.from_numpy(smsk).cuda())
bdata.append(torch.from_numpy(imgt).cuda())
bdata.append(torch.from_numpy(tmsk).cuda())
with torch.no_grad():
boxout = boxnet(bdata)
#
size = np.sum(np.sum(bdata[1].data.cpu().numpy() > 0, axis=1),
axis=1)
print(size)
undone_queue = []
for idx, v in enumerate(size):
heapq.heappush(undone_queue, (-v, idx))
done_queue = []
msk_in = []
box_out = []
box_color = []
box_gt = []
s_gt = []
s_out = []
t_gt = []
t_out = []
r1_gt = []
r1_out = []
r2_gt = []
r2_out = []
baset = np.zeros([len(box_lst), 3], dtype=np.float32)
bases = np.zeros([len(box_lst), 1], dtype=np.float32)
while len(undone_queue) > 0:
if len(done_queue) > 0:
itop = heapq.heappop(done_queue)
ci = itop[1]
bt = baset[ci, :][np.newaxis, :]
print('[done:%d' % ci, file=logf, end='')
else:
itop = heapq.heappop(undone_queue)
ci = itop[1]
bo = boxout['sb'].data.cpu().numpy()[ci, ...]
bgt, rs, t = parsegt(box_lst[ci])
add_gt_for_eval(box_lst[ci], s_gt, t_gt, r1_gt, r2_gt)
add_out_for_eval(
boxout['ss'].data[ci, ...] * rs,
torch.from_numpy(t.astype(np.float32)).cuda(),
boxout['sr1'].data[ci, ...], boxout['sr2'].data[ci,
...],
s_out, t_out, r1_out, r2_out)
baset[ci, :] = t[0, :]
bases[ci, :] = rs
bt = t
box_out.append(bo * rs + bt)
box_gt.append(bgt)
box_color.append(red_box)
msk_in.append(msk_lst[ci])
print('[undone:%d' % ci, file=logf, end='')
unfinish = []
while len(undone_queue) > 0:
tdata = []
tptdata = []
jtop = heapq.heappop(undone_queue)
cj = jtop[1]
pimg, psmsk, ptmsk = msk_pair_center(
img, msk_lst[ci], msk_lst[cj])
pimg = torch.from_numpy(pimg).cuda().unsqueeze(0)
psmsk = torch.from_numpy(psmsk).cuda().unsqueeze(0)
ptmsk = torch.from_numpy(ptmsk).cuda().unsqueeze(0)
tdata.append(pimg)
tdata.append(psmsk)
tdata.append(ptmsk)
with torch.no_grad():
touchout = touchnet(tdata)
if touchout['y'].data.cpu().numpy()[0][0] > 0.5:
im = Image.fromarray((pimg.data.cpu().numpy()[0, ...] *
255).astype(np.uint8))
im.save(
os.path.join(fopath,
'_003_%03d_img.png' % (len(box_out))))
allmsk = (psmsk + ptmsk).data.cpu().numpy()[0, ...]
imt = Image.fromarray((allmsk * 255).astype(np.uint8))
imt.save(
os.path.join(fopath,
'_004_%03d_msk.png' % (len(box_out))))
bgt, rsgt, _ = parsegt(box_lst[cj])
tptdata.append(pimg)
tptdata.append(psmsk)
tptdata.append(ptmsk)
tptdata.append(None)
vec = np.zeros([1, 21], dtype=np.float32)
rs = bases[ci, :]
vec[0, :3] = rs * boxout['ss'].data.cpu().numpy()[ci,
...]
vec[0, 3:6] = boxout['sr1'].data.cpu().numpy()[ci, ...]
vec[0, 6:9] = boxout['sr2'].data.cpu().numpy()[ci, ...]
vec[0,
9:12] = rs * boxout['ss'].data.cpu().numpy()[cj,
...]
vec[0, 15:18] = boxout['sr1'].data.cpu().numpy()[cj,
...]
vec[0, 18:21] = boxout['sr2'].data.cpu().numpy()[cj,
...]
tptdata.append(torch.from_numpy(vec).cuda())
with torch.no_grad():
tptout = touchptnet(tptdata)
heapq.heappush(done_queue, jtop)
t = tptout['t'].data.cpu().numpy()
bo = tptout['tb'].data.cpu().numpy()[0, ...]
box_out.append(bo + bt)
baset[cj, :] = (t + bt)[0, :]
sout = boxout['ss'].data.cpu().numpy()[
cj, ...] * tptout['ts'].data.cpu().numpy()[0, ...]
bases[cj, :] = np.max(np.abs(sout))
box_gt.append(bgt)
box_color.append(blue_box)
msk_in.append(bdata[1].data.cpu().numpy()[cj, ...])
add_gt_for_eval(box_lst[cj], s_gt, t_gt, r1_gt, r2_gt)
add_out_for_eval(
boxout['ss'].data[cj, ...] *
tptout['ts'].data[0, ...],
torch.from_numpy(
(t + bt).astype(np.float32)).cuda(),
boxout['sr1'].data[cj,
...], boxout['sr2'].data[cj,
...],
s_out, t_out, r1_out, r2_out)
print('->%d' % cj, file=logf, end='')
else:
unfinish.append(jtop)
for p in unfinish:
heapq.heappush(undone_queue, p)
print(']', file=logf)
im = Image.fromarray((img * 255).astype(np.uint8))
im.save(os.path.join(fopath, '_001_000_im.png'))
writegt(fopath, box_gt)
writeout(fopath, box_out, box_color, msk_in)
val = eval(s_out, t_out, r1_out, r2_out, s_gt, t_gt, r1_gt, r2_gt)
v = float(val.data.cpu().numpy())
json.dump({'bcd': v}, open(os.path.join(fopath, 'meta.json'), 'w'))
if cat != 'TrainChair':
eval_all_sum += val.data.cpu().numpy()
eval_all_cnt += 1.0
eval_cat_sum += val.data.cpu().numpy()
eval_cat_cnt += 1.0
if cat != 'TrainChair':
print(cat + ',%f' % (eval_cat_sum / eval_cat_cnt), file=eval_log)
eval_log.flush()
print('inst mean,%f' % (eval_all_sum / eval_all_cnt), file=eval_log)
eval_log.close()
def run(**kwargs):
global iternum;
#get configuration
try:
config = importlib.import_module('config.'+kwargs['config']);
opt = config.__dict__;
for k in kwargs.keys():
if not kwargs[k] is None:
opt[k] = kwargs[k];
except Exception as e:
print(e);
traceback.print_exc();
exit();
#get network
try:
m = importlib.import_module('net.'+opt['touch_net']);
touchnet = m.Net(**opt);
#
m = importlib.import_module('net.'+opt['box_net']);
boxnet = m.Net(**opt);
#
m = importlib.import_module('net.'+opt['touchpt_net']);
touchptnet = m.Net(**opt);
#
if torch.cuda.is_available():
touchnet = touchnet.cuda();
touchnet.eval();
boxnet = boxnet.cuda();
boxnet.eval();
touchptnet = touchptnet.cuda();
touchptnet.eval();
#
partial_restore(touchnet,opt['touch_model']);
partial_restore(boxnet,opt['box_model']);
partial_restore(touchptnet,opt['touchpt_model']);
except Exception as e:
print(e);
traceback.print_exc();
exit();
#get dataset
dpath = opt['data_path'];
im_lst = os.listdir(dpath);
for im in im_lst:
if im.endswith('_im.png'):
image = np.array(Image.open(os.path.join(dpath,im))).astype(np.float32) / 255.0;
bdata = [];
img_lst = [];
nm_lst = [];
smsk_lst = [];
tmsk_lst = [];
for msk in im_lst:
if msk.endswith('_msk.png') and (im.split('_')[0] in msk):
nm_lst.append(msk);
img_lst.append(image);
mskimg = np.array(Image.open(os.path.join(dpath,msk))).astype(np.float32) / 255.0;
smsk_lst.append(mskimg);
tmsk_lst.append(mskimg);
print(mskimg.shape);
img = np.stack(img_lst,axis=0);
smsk = np.stack(smsk_lst,axis=0);
tmsk = np.stack(tmsk_lst,axis=0);
bdata.append(torch.from_numpy(img).cuda());
bdata.append(torch.from_numpy(smsk).cuda());
bdata.append(torch.from_numpy(tmsk).cuda());
with torch.no_grad():
boxout = boxnet(bdata);
bx = boxout['sb'].data.cpu().numpy();
for i,nm in enumerate(nm_lst):
writebox(os.path.join(dpath,nm.replace('_msk.png','_box.ply')),[bx[i,...]]);