def application_openFile_(self, app, f):
extension = os.path.splitext(f)[1].lower()
if extension not in ('.nno', '.json', '.cadnano'):
print "Could not open file %s (bad extension %s)"%(f, extension)
return
doc = decode(file(str(f)).read())
DocumentController(doc, str(f))
return None
def application_openFile_(self, app, f):
if f == "main.py": # ignore
return
extension = os.path.splitext(f)[1].lower()
if extension not in ('.nno', '.json', '.cadnano'):
print "Could not open file %s (bad extension %s)" % (f, extension)
return
doc = decode(file(str(f)).read())
DocumentController(doc, str(f))
return None
def application_openFile_(self, app, f):
if f == "main.py": # ignore
return
extension = os.path.splitext(f)[1].lower()
if extension not in (".nno", ".json", ".cadnano"):
print "Could not open file %s (bad extension %s)" % (f, extension)
return
doc = decode(file(str(f)).read())
DocumentController(doc, str(f))
return None
def newDocument(self, isFirstNewDoc=False):
from controllers.documentcontroller import DocumentController
defaultFile = environ.get('CADNANO_DEFAULT_DOCUMENT', None)
if defaultFile and isFirstNewDoc:
defaultFile = path.expanduser(defaultFile)
defaultFile = path.expandvars(defaultFile)
from model.decoder import decode
doc = decode(file(defaultFile).read())
print "Loaded default document: %s" % doc
dc = DocumentController(doc, defaultFile)
else:
dc = DocumentController() # DocumentController is responsible
# for adding itself to
# app.documentControllers
return dc.document()
def openFile(self, selected):
if isinstance(selected, QStringList) or isinstance(selected, list):
fname = selected[0]
else:
fname = selected
if not fname or os.path.isdir(fname):
return False
fname = str(fname)
doc = decode(file(fname).read())
doc.finalizeImport() # updates staple highlighting
DocumentController(doc, fname)
if self.filesavedialog != None:
self.filesavedialog.filesSelected.disconnect(self.openFile)
# manual garbage collection to prevent hang (in osx)
del self.filesavedialog
def newDocument(self, isFirstNewDoc=False):
from controllers.documentcontroller import DocumentController
defaultFile = environ.get('CADNANO_DEFAULT_DOCUMENT', None)
if defaultFile and isFirstNewDoc:
defaultFile = path.expanduser(defaultFile)
defaultFile = path.expandvars(defaultFile)
from model.decoder import decode
doc = decode(file(defaultFile).read())
print "Loaded default document: %s" % doc
dc = DocumentController(doc, defaultFile)
else:
docCtrlrCount = len(self.documentControllers)
if docCtrlrCount == 0: # first dc
# dc adds itself to app.documentControllers
dc = DocumentController()
elif docCtrlrCount == 1: # dc already exists
dc = list(self.documentControllers)[0]
dc.newDocument() # tell it to make a new doucment
return dc.document()
def test():
print("testing...")
Config.batch_size = 1
Config.model_path = "testmodel_last.pt"
print("mAP files output path: " + Config.mAP_path)
model_path = Config.model_path
print("model: ", model_path)
print("conf: ", Config.conf_thresh)
print("iou thresh: ", Config.conf_thresh)
if Config.dataset == "VOC":
print("dataset: ", Config.voc_dataset_dir)
dataset = DatasetJNN_VOC(Config.voc_dataset_dir,
mode="test",
year="2007",
is_training=False)
elif Config.dataset == "coco":
print("dataset: ", Config.coco_dataset_dir)
dataset = DatasetJNN_COCO(Config.coco_dataset_dir,
is_training=False)
else:
print("dataset: ", Config.testing_dir)
folder_dataset = dset.ImageFolder(root=Config.testing_dir)
dataset = DatasetJNN(imageFolderDataset=folder_dataset,
is_training=False)
dataloader = DataLoader(dataset,
shuffle=False,
num_workers=0,
batch_size=1)
model = DarkJNN()
checkpoint = torch.load(model_path)
model.load_state_dict(checkpoint['model'])
model.cuda()
model.eval()
with torch.no_grad():
for i, data in enumerate(dataloader, 0):
if (i % 1000 == 0):
print(str(i) + "/" + str(len(dataset))) # progress
img0, img1, targets, label, im_infos = data
img0, img1, targets = Variable(img0).cuda(), Variable(
img1).cuda(), targets.cuda()
model_output = model(img0, img1, targets)
im_info = {
'width': im_infos[0].item(),
'height': im_infos[1].item()
}
output = [item[0].data for item in model_output]
detections = decode(output,
im_info,
conf_threshold=Config.conf_thresh,
nms_threshold=Config.nms_thresh)
if len(detections) > 0:
# mAP files
pair_id = im_infos[2][0].split(
'.')[0] + "_" + im_infos[3][0].split('.')[0]
detection_str = ""
gt_str = ""
f = open(
Config.mAP_path + "groundtruths/" + pair_id + ".txt",
"a+")
for box_idx in range(len(targets)):
gt_str += label[0] + " " \
+ str(targets[0][box_idx][0].item()) + " " \
+ str(targets[0][box_idx][1].item()) + " " \
+ str(targets[0][box_idx][2].item()) + " " \
+ str(targets[0][box_idx][3].item()) + "\n"
f.seek(0)
if not (gt_str in f.readlines()):
f.write(gt_str)
f.close()
f = open(
Config.mAP_path + "detections/" + pair_id + ".txt",
"a+")
for detection in detections:
detection_str += label[0] + " " \
+ str(detection[4].item()) + " "\
+ str(detection[0].item()) + " "\
+ str(detection[1].item()) + " "\
+ str(detection[2].item()) + " "\
+ str(detection[3].item()) + "\n"
f.seek(0)
if not (detection_str in f.readlines()):
f.write(detection_str)
f.close()
def test_one_COCO():
""" Tests a a pair of images """
print("testing one image...")
Config.model_path = "/home/mmv/Documents/2.projects/JNN_detection/trained_models/dJNN_COCOsplit2/testmodel_last_split2.pt"
model_path = Config.model_path
model = DarkJNN()
checkpoint = torch.load(model_path)
model.load_state_dict(checkpoint['model'])
model.cuda()
model.eval()
coco_dataset = dset.CocoDetection(
Config.coco_dataset_dir,
Config.coco_dataset_dir + "annotations/instances_val2017.json")
# (3m1, 3m6), (rbc1, rbc43), hp(33971473, 70609284), blizzard(1, 6), gen_electric(7, 31), warner(10, 18)
# goodyear(13, 20), airhawk(12, 1), gap(34, 36), levis(14, 30)
q_name = "000000024144"
t_name = "000000306700"
q_im = Image.open("/home/mmv/Documents/3.datasets/coco/val2017/" +
q_name + ".jpg")
t_im = Image.open("/home/mmv/Documents/3.datasets/coco/val2017/" +
t_name + ".jpg")
# find image id and (first) annotation
for id in coco_dataset.coco.imgs:
if coco_dataset.coco.imgs[id]['file_name'] == q_name + ".jpg":
break
for ann_id in coco_dataset.coco.anns:
if coco_dataset.coco.anns[ann_id]['image_id'] == id:
print(coco_dataset.coco.anns[ann_id])
break
qbox = coco_dataset.coco.anns[ann_id]['bbox']
qbox = [qbox[0], qbox[1], qbox[0] + qbox[2], qbox[1] + qbox[3]]
q_im = q_im.crop((qbox[0], qbox[1], qbox[2], qbox[3]))
w, h = t_im.size[0], t_im.size[1]
im_infos = (w, h, q_name, t_name)
qcv_im = np.array(q_im)
qcv_im = qcv_im[:, :, ::-1].copy()
cv_im = np.array(t_im)
cv_im = cv_im[:, :, ::-1].copy()
q_im = q_im.resize((Config.imq_w, Config.imq_h))
t_im = t_im.resize((Config.im_w, Config.im_h))
# To float tensors
q_im = torch.from_numpy(np.array(q_im)).float() / 255
t_im = torch.from_numpy(np.array(t_im)).float() / 255
img0 = q_im.permute(2, 0, 1)
img1 = t_im.permute(2, 0, 1)
img0 = torch.unsqueeze(img0, 0)
img1 = torch.unsqueeze(img1, 0)
with torch.no_grad():
#
img0, img1 = Variable(img0).cuda(), Variable(img1).cuda()
model_output = model(img0, img1, [])
im_info = {'width': im_infos[0], 'height': im_infos[1]}
output = [item[0].data for item in model_output]
detections = decode(output,
im_info,
conf_threshold=Config.conf_thresh,
nms_threshold=Config.nms_thresh)
if len(detections) > 0:
for detection in detections:
start_pt = (int(detection[0].item()),
int(detection[1].item()))
end_pt = (int(detection[2].item()),
int(detection[3].item()))
image = cv2.rectangle(cv_im, start_pt, end_pt, (0, 255, 0),
3)
print(start_pt, end_pt)
cv2.imshow("q", qcv_im)
cv2.imshow("res", image)
cv2.waitKey()
else:
print("No detctions found")
def test_one_OL():
""" Tests a a pair of images """
print("testing one image...")
Config.model_path = "/home/mmv/Documents/2.projects/JNN_detection/trained_models/dJNN_COCOsplit2/testmodel_last_split2.pt"
model_path = Config.model_path
model = DarkJNN()
checkpoint = torch.load(model_path)
model.load_state_dict(checkpoint['model'])
model.cuda()
model.eval()
# (3m1, 3m6), (rbc1, rbc43), hp(33971473, 70609284), blizzard(1, 6), gen_electric(7, 31), warner(10, 18)
# goodyear(13, 20), airhawk(12, 1), gap(34, 36), levis(14, 30)
q_name = "000000008629"
t_name = "000000209530"
q_im = Image.open("/home/mmv/Documents/3.datasets/coco/val2017/" +
q_name + ".jpg")
t_im = Image.open("/home/mmv/Documents/3.datasets/coco/val2017/" +
t_name + ".jpg")
w, h = t_im.size[0], t_im.size[1]
im_infos = (w, h, q_name, t_name)
cv_im = np.array(t_im)
cv_im = cv_im[:, :, ::-1].copy()
q_im = q_im.resize((Config.imq_w, Config.imq_h))
t_im = t_im.resize((Config.im_w, Config.im_h))
# To float tensors
q_im = torch.from_numpy(np.array(q_im)).float() / 255
t_im = torch.from_numpy(np.array(t_im)).float() / 255
img0 = q_im.permute(2, 0, 1)
img1 = t_im.permute(2, 0, 1)
img0 = torch.unsqueeze(img0, 0)
img1 = torch.unsqueeze(img1, 0)
with torch.no_grad():
#
img0, img1 = Variable(img0).cuda(), Variable(img1).cuda()
model_output = model(img0, img1, [])
im_info = {'width': im_infos[0], 'height': im_infos[1]}
output = [item[0].data for item in model_output]
detections = decode(output,
im_info,
conf_threshold=Config.conf_thresh,
nms_threshold=Config.nms_thresh)
if len(detections) > 0:
for detection in detections:
start_pt = (int(detection[0].item()),
int(detection[1].item()))
end_pt = (int(detection[2].item()),
int(detection[3].item()))
image = cv2.rectangle(cv_im, start_pt, end_pt, (0, 255, 0),
3)
print(start_pt, end_pt)
cv2.imshow("res", image)
cv2.waitKey()