roi = [slice(652319, 652637)]; # turn #roi = [slice(586000, 600000)]; # no shape detection failure images = data_images[roi]; shapes = data_shapes[roi]; shape_info = data_shape_info[roi]; nt = shapes.shape[0]; #%% blur = np.zeros(images.shape); for i in range(nt): blur[i] = cv2.GaussianBlur(images[i], ksize = (5,5), sigmaX = 0); dv.plot(blur); #%% head tail trajectories head_tails = []; for i in range(nt): center = shapes[i][0:2].T; head_tails.append(center[[0,-1]]); head_tails = np.array(head_tails); #%% plt.figure(1); plt.clf(); plt.scatter(*(head_tails[:,0,:].T), color = 'red')
r = perimeter / 2 / np.pi
a = np.pi * r**2
print a, area
#%% cureled
#%%
blur2 = np.zeros(turn2.shape)
for i in range(blur2.shape[0]):
blur2[i] = cv2.GaussianBlur(np.asarray(turn2[i], dtype=float),
ksize=(5, 5),
sigmaX=0)
dv.plot(blur2)
#%%
plt.figure(7)
plt.clf()
for d in range(3):
plt.subplot(1, 3, d + 1)
plt.imshow(frame_color[:, :, d])
#%%
dv.plot(data_images.transpose([1, 2, 0]))
#%%
fids = range(500000, 500000+300);
ht = np.array([data_shape[f][:2,[0,-1]].T for f in fids])
plt.figure(16); plt.clf();
plt.scatter(ht[:,0,0], ht[:,0,1], c = 'r')
plt.scatter(ht[:,1,0], ht[:,1,1], c = 'b')
#%%
d = data_image[fids]
dv.plot(d.transpose([1,2,0]));
#%% draw contours on images and thenplot via dv
fids = [9];
fids = range(520000+400, 520000+800);
n = len(fids);
dc = np.zeros((n, data_image.shape[1], data_image.shape[2]));
for i,f in enumerate(fids):
img = np.asarray(data_image[f].copy(), dtype = 'float32');
img = cv2.GaussianBlur(img, ksize = (3,3), sigmaX = 0);
cnt = data_shape[f][:2].T;
cnt = np.asarray(np.reshape(cnt, (cnt.shape[0], 1, cnt.shape[1])), dtype = int);
dc[i] = cv2.polylines(img, [cnt], False, 255, 1);
data_meta_file = os.path.join(data_dir, data_name % (region_id, 'meta'))
data_shape_file = os.path.join(data_dir, data_name % (region_id, 'shapes'))
#%%
turn = np.load(os.path.join(data_dir, 'Tests/turn_000.npy'))
#dv.plot(turn)
data = -turn
blur = np.zeros(data.shape)
for i in range(data.shape[0]):
blur[i] = cv2.GaussianBlur(data[i], ksize=(5, 5), sigmaX=0)
dv.plot(blur > 185)
#%%
data0 = blur * (blur > 185)
for i in range(data0.shape[0]):
data0[i][data0[i] > 0] = i + 1
#%%
import vispy as vp
#from vispy import app, visuals, scene
# build visuals
GraphPlot3D = vp.scene.visuals.create_visual_node(vp.visuals.VolumeVisual)
fids = range(500000, 500000+300);
ht = np.array([data_shape[f][:2,[0,-1]].T for f in fids])
plt.figure(16); plt.clf();
plt.scatter(ht[:,0,0], ht[:,0,1], c = 'r')
plt.scatter(ht[:,1,0], ht[:,1,1], c = 'b')
#%%
d = data_image[fids]
dv.plot(d.transpose([1,2,0]));
#%% draw contours on images and thenplot via dv
fids = [9];
fids = range(520000+400, 520000+800);
n = len(fids);
dc = np.zeros((n, data_image.shape[1], data_image.shape[2]));
for i,f in enumerate(fids):
img = np.asarray(data_image[f].copy(), dtype = 'float32');
img = cv2.GaussianBlur(img, ksize = (3,3), sigmaX = 0);
cnt = data_shape[f][:2].T;
cnt = np.asarray(np.reshape(cnt, (cnt.shape[0], 1, cnt.shape[1])), dtype = int);
dc[i] = cv2.polylines(img, [cnt], False, 255, 1);
#%% pmu.get_movie_from_frame(movie_files, failed_id[0], n_frames=n_frames) #%% plt.figure(1); plt.clf(); plt.imshow(data_image[263332-37443]) #%% dv.plot(data_image.transpose([2,1,0])) #%% curled 263332-37520 #%% turns i = 337714; i = 109715; turn = data_image[i-75:i+150]; dv.plot(turn)
n_frames = pmu.get_n_frames(movie_files) #%% pmu.get_movie_from_frame(movie_files, failed_id[0], n_frames=n_frames) #%% plt.figure(1) plt.clf() plt.imshow(data_image[263332 - 37443]) #%% dv.plot(data_image.transpose([2, 1, 0])) #%% curled 263332 - 37520 #%% turns i = 337714 i = 109715 turn = data_image[i - 75:i + 150] dv.plot(turn) #%%
#%% cureled #%% blur2 = np.zeros(turn2.shape); for i in range(blur2.shape[0]): blur2[i] = cv2.GaussianBlur(np.asarray(turn2[i], dtype = float), ksize = (5,5), sigmaX = 0); dv.plot(blur2); #%% plt.figure(7); plt.clf(); for d in range(3): plt.subplot(1,3,d+1); plt.imshow(frame_color[:,:,d]) #%% dv.plot(data_images.transpose([1,2,0]));
d = cv2.GaussianBlur(np.asarray(data_image[i], dtype = 'float'), ksize= (3,3), sigmaX = 0);
dz = d[75-w:75+w+1, 75-w:75+w+1].flatten();
dz[dz < 135] = 135;
dz[dz > 145] = 145;
features[j] = np.dot(dz, project);
features = (features - np.mean(features, axis = 0));
features = (features / np.std(features, axis = 0));
print features.shape
#%%
dv.plot(features)
#%%
import sklearn.manifold as sl;
n_components = 2;
metric = 'euclidean';
tsne = sl.TSNE(n_components=n_components, init = 'pca', random_state = 10, metric = metric)
Y = tsne.fit_transform(features)
#%%
fig = plt.figure(17);
plt.clf();
plt.scatter(Y[:,0], Y[:,1], c = range(len(Y[:,0])), cmap = plt.cm.Spectral);
plt.title("t-SNE")
data_shape_file = os.path.join(data_dir, data_name % (region_id, 'shapes')); #%% turn = np.load(os.path.join(data_dir, 'Tests/turn_000.npy')) #dv.plot(turn) data = -turn; blur = np.zeros(data.shape); for i in range(data.shape[0]): blur[i] = cv2.GaussianBlur(data[i], ksize = (5,5), sigmaX = 0); dv.plot(blur>185) #%% data0 = blur * (blur>185); for i in range(data0.shape[0]): data0[i][data0[i] > 0] = i +1; #%% import vispy as vp; #from vispy import app, visuals, scene # build visuals GraphPlot3D = vp.scene.visuals.create_visual_node(vp.visuals.VolumeVisual)