python类COLORMAP_JET的实例源码

def create_heatmaps(img, pred):
    """
    Uses objectness probability to draw a heatmap on the image and returns it
    """
    # find anchors with highest prediction
    best_pred = np.max(pred[..., 0], axis=-1)
    # convert probabilities to colormap scale
    best_pred = np.uint8(best_pred * 255)
    # apply color map
    # cv2 colormaps create BGR, not RGB
    cmap = cv2.cvtColor(cv2.applyColorMap(best_pred, cv2.COLORMAP_JET), cv2.COLOR_BGR2RGB)
    # resize the color map to fit image
    cmap = cv2.resize(cmap, img.shape[1::-1], interpolation=cv2.INTER_NEAREST)

    # overlay cmap with image
    return cv2.addWeighted(cmap, 1, img, 0.5, 0)

def save(mask, img, blurred):
    mask = mask.cpu().data.numpy()[0]
    mask = np.transpose(mask, (1, 2, 0))

    mask = (mask - np.min(mask)) / np.max(mask)
    mask = 1 - mask
    heatmap = cv2.applyColorMap(np.uint8(255*mask), cv2.COLORMAP_JET)

    heatmap = np.float32(heatmap) / 255
    cam = 1.0*heatmap + np.float32(img)/255
    cam = cam / np.max(cam)

    img = np.float32(img) / 255
    perturbated = np.multiply(1 - mask, img) + np.multiply(mask, blurred)   

    cv2.imwrite("perturbated.png", np.uint8(255*perturbated))
    cv2.imwrite("heatmap.png", np.uint8(255*heatmap))
    cv2.imwrite("mask.png", np.uint8(255*mask))
    cv2.imwrite("cam.png", np.uint8(255*cam))

def likelihood_map(prob_map,image) :
    '''This functon generates the likelihood map based on either obj-surr/dist model
       input: probability map
       output:likelihood map, an image(each pixel value=corresponding probability)'''

    global h_img,w_img,bin
    sf=256.0/bin
    image_10=image/sf 
    image_10=image_10.astype('uint8')
    # creating a likelihood image acc. to obj-surr or obj-distractor model
    a=image_10[:,:,0]
    a=a.ravel()
    b=image_10[:,:,1]
    b=b.ravel()
    c_=image_10[:,:,2]
    c_=c_.ravel()
    prob_image=prob_map[a,b,c_]
    prob_image=prob_image.reshape((h_img,w_img))
    prob_image1=prob_image*255
    prob_image1=prob_image1.astype('uint8')
    likemap=cv2.applyColorMap(prob_image1, cv2.COLORMAP_JET)
    return likemap,prob_image1

def likelihood_map(prob_map,image) :
    '''This functon generates the likelihood map based on either obj-surr/dist model
       input: probability map
       output:likelihood map, an image(each pixel value=corresponding probability)'''

    global h_img,w_img,bin
    sf=256.0/bin
    image_10=image/sf 
    image_10=image_10.astype('uint8')
    # creating a likelihood image acc. to obj-surr or obj-distractor model
    a=image_10[:,:,0]
    a=a.ravel()
    b=image_10[:,:,1]
    b=b.ravel()
    c_=image_10[:,:,2]
    c_=c_.ravel()
    prob_image=prob_map[a,b,c_]
    prob_image=prob_image.reshape((h_img,w_img))
    prob_image1=prob_image*255
    prob_image1=prob_image1.astype('uint8')
    likemap=cv2.applyColorMap(prob_image1, cv2.COLORMAP_JET)
    return likemap,prob_image1

def visualize_hypercolumns(model, original_img):

    img = np.float32(cv2.resize(original_img, (200, 66))) / 255.0

    layers_extract = [9]

    hc = extract_hypercolumns(model, layers_extract, img)
    avg = np.product(hc, axis=0)
    avg = np.abs(avg)
    avg = avg / np.max(np.max(avg))

    heatmap = cv2.applyColorMap(np.uint8(255 * avg), cv2.COLORMAP_JET)
    heatmap = np.float32(heatmap) / np.max(np.max(heatmap))
    heatmap = cv2.resize(heatmap, original_img.shape[0:2][::-1])

    both = 255 * heatmap * 0.7 + original_img
    both = both / np.max(both)
    return both

def recalculate(self):

        in_mark = self.g_pool.trim_marks.in_mark
        out_mark = self.g_pool.trim_marks.out_mark
        section = slice(in_mark,out_mark)

        # calc heatmaps
        for s in self.surfaces:
            if s.defined:
                s.generate_heatmap(section)

        # calc distirbution accross all surfaces.
        results = []
        for s in self.surfaces:
            gaze_on_srf  = s.gaze_on_srf_in_section(section)
            results.append(len(gaze_on_srf))
            self.metrics_gazecount = len(gaze_on_srf)

        if results == []:
            logger.warning("No surfaces defined.")
            return
        max_res = max(results)
        results = np.array(results,dtype=np.float32)
        if not max_res:
            logger.warning("No gaze on any surface for this section!")
        else:
            results *= 255./max_res
        results = np.uint8(results)
        results_c_maps = cv2.applyColorMap(results, cv2.COLORMAP_JET)

        for s,c_map in zip(self.surfaces,results_c_maps):
            heatmap = np.ones((1,1,4),dtype=np.uint8)*125
            heatmap[:,:,:3] = c_map
            s.metrics_texture = Named_Texture()
            s.metrics_texture.update_from_ndarray(heatmap)

def save_cam_image(img, mask, filename):
    heatmap = cv2.applyColorMap(np.uint8(255*mask), cv2.COLORMAP_JET)
    heatmap = np.float32(heatmap) / 255
    cam = heatmap + np.float32(img)
    cam = cam / np.max(cam)
    cv2.imwrite(filename, np.uint8(255 * cam))

def opencv_plot(des_name):
    densmap = np.fromfile(densmap_name, np.float32)
    densmap = np.reshape(densmap, (227, 227))
    #densmap = norm_image(densmap) * 100 
    densmap *= 100.0
    densmap[densmap >1 ] = 1
    densmap = norm_image(densmap) * 255
    densmap = densmap.astype(np.uint8)
    im_color = cv2.applyColorMap(densmap, cv2.COLORMAP_JET)
    cv2.imshow("im", im_color)
    cv2.waitKey(0)

def predict_image(flag):
    t_start = cv2.getTickCount()
    config = tf.ConfigProto()
    # config.gpu_options.per_process_gpu_memory_fraction = 0.9
    config.gpu_options.allow_growth = True
    set_session(tf.Session(config=config))

    with open(os.path.join(flag.ckpt_dir, flag.ckpt_name, 'model.json'), 'r') as json_file:
            loaded_model_json = json_file.read()
    model = model_from_json(loaded_model_json)
    weight_list = sorted(glob(os.path.join(flag.ckpt_dir, flag.ckpt_name, "weight*")))
    model.load_weights(weight_list[-1])
    print "[*] model load : %s"%weight_list[-1]
    t_total = (cv2.getTickCount() - t_start) / cv2.getTickFrequency() * 1000 
    print "[*] model loading Time: %.3f ms"%t_total

    imgInput = cv2.imread(flag.test_image_path, 0)
    input_data = imgInput.reshape((1,256,256,1))

    t_start = cv2.getTickCount()
    result = model.predict(input_data, 1)
    t_total = (cv2.getTickCount() - t_start) / cv2.getTickFrequency() * 1000
    print "Predict Time: %.3f ms"%t_total

    imgMask = (result[0]*255).astype(np.uint8)
    imgShow = cv2.cvtColor(imgInput, cv2.COLOR_GRAY2BGR)
    _, imgMask = cv2.threshold(imgMask, int(255*flag.confidence_value), 255, cv2.THRESH_BINARY)
    imgMaskColor = cv2.applyColorMap(imgMask, cv2.COLORMAP_JET)
    # imgZero = np.zeros((256,256), np.uint8)
    # imgMaskColor = cv2.merge((imgZero, imgMask, imgMask))
    imgShow = cv2.addWeighted(imgShow, 0.9, imgMaskColor, 0.3, 0.0)
    output_path = os.path.join(flag.output_dir, os.path.basename(flag.test_image_path))
    cv2.imwrite(output_path, imgShow)
    print "SAVE:[%s]"%output_path

def train_visualization_seg(self, model, epoch):
        image_name_list = sorted(glob(os.path.join(self.flag.data_path,'train/IMAGE/*/*.png')))
        print image_name_list

        image_name = image_name_list[-1]
        image_size = self.flag.image_size

        imgInput = cv2.imread(image_name, self.flag.color_mode)
        output_path = self.flag.output_dir
        input_data = imgInput.reshape((1,image_size,image_size,self.flag.color_mode*2+1))

        t_start = cv2.getTickCount()
        result = model.predict(input_data, 1)
        t_total = (cv2.getTickCount() - t_start) / cv2.getTickFrequency() * 1000
        print "[*] Predict Time: %.3f ms"%t_total

        imgMask = (result[0]*255).astype(np.uint8)
        imgShow = cv2.cvtColor(imgInput, cv2.COLOR_GRAY2BGR)
        imgMaskColor = cv2.applyColorMap(imgMask, cv2.COLORMAP_JET)
        imgShow = cv2.addWeighted(imgShow, 0.9, imgMaskColor, 0.4, 0.0)
        output_path = os.path.join(self.flag.output_dir, '%04d_'%epoch+os.path.basename(image_name))
        cv2.imwrite(output_path, imgShow)
        # print "SAVE:[%s]"%output_path
        # cv2.imwrite(os.path.join(output_path, 'img%04d.png'%epoch), imgShow)
        # cv2.namedWindow("show", 0)
        # cv2.resizeWindow("show", 800, 800)
        # cv2.imshow("show", imgShow)
        # cv2.waitKey(1)

def grad_cam(input_model, image, weights, feature_maps=None):
    #activation size of final convolition layer is 10x10"
    cam = np.ones((10, 10), dtype=np.float32)
    # Add weighted activation maps
    grads_val = weights
    for i in range(grads_val.shape[0]):
        # Added relu
        temp = (weights[i, :] * feature_maps[:, :, i])
        np.maximum(temp, 0, temp)
        cam += temp

    # resize and normalization
    del feature_maps
    cam = cv2.resize(cam, (299, 299))
    # Relu
    cam = np.maximum(cam, 0)
    cam = cam / np.max(cam)
    image = image[0, :]
    image -= np.min(image)
    image = np.minimum(image, 255)
   # print image.shape


    cam = cv2.applyColorMap(np.uint8(255 * cam), cv2.COLORMAP_JET)   
    cam = 0.5*np.float32(cam) + 0.5*np.float32(image)
    cam = 255 * cam / np.max(cam)
    return np.uint8(cam)

def save(self, filename, gcam, raw_image):
        gcam = cv2.applyColorMap(np.uint8(gcam * 255.0), cv2.COLORMAP_JET)
        gcam = gcam.astype(np.float) + raw_image.astype(np.float)
        gcam = gcam / gcam.max() * 255.0
        cv2.imwrite(filename, np.uint8(gcam))

def visualize_grad_cam(input_model, original_img, layer_name = "conv3_1"): 

    img = np.float32(cv2.resize(original_img, (200, 66))) / 255.0

    angle = input_model.predict(np.array([img]))
    print("The predicted angle is", 180.0 * angle[0][0] / scipy.pi, "degrees")

    model = Sequential()
    model.add(input_model)

    target_layer = lambda x: grad_cam_loss(x, angle)
    model.add(Lambda(target_layer,
                     output_shape = grad_cam_loss_output_shape))

    loss = K.sum(model.layers[-1].output)
    conv_output =  [l for l in model.layers[0].layers if l.name is layer_name][0].output
    grads = normalize(K.gradients(loss, conv_output)[0])
    gradient_function = K.function([model.layers[0].input], [conv_output, grads])

    output, grads_val = gradient_function([[img]])
    output, grads_val = output[0, :], grads_val[0, :, :, :]

    weights = np.mean(grads_val, axis = (0, 1))
    cam = np.ones(output.shape[0 : 2], dtype = np.float32)

    for i, w in enumerate(weights):
        cam += w * output[:, :, i]

    #ReLU:
    cam = np.maximum(cam, 0)
    cam = cam / np.max(cam)
    cam = cv2.resize(cam, tuple(original_img.shape[0:2][::-1]))

    cam = cv2.applyColorMap(np.uint8(255*cam), cv2.COLORMAP_JET)

    cam = 1.0 * np.float32(cam) + np.float32(original_img)
    cam = cam / np.max(cam)
    return cam

def showImg(self,label,img):
        if len(img.shape) == 2:
            img = cv2.applyColorMap(img, cv2.COLORMAP_JET)
            img = cv2.resize(img, (512, 512),cv2.INTER_AREA)
        height, width, byteValue = img.shape
        byteValue = byteValue * width
        timg = img.copy()
        image = QtGui.QImage(timg.data, width, height,byteValue, QtGui.QImage.Format_RGB888)
        label.setPixmap(QtGui.QPixmap(image).scaled(label.size(),aspectMode=QtCore.Qt.KeepAspectRatio))     


        """ visualize function from
        https://github.com/BVLC/caffe/blob/master/examples/00-classification.ipynb
        """

def vis_square(self, data):
        """Take an array of shape (n, height, width) or (n, height, width, 3)
           and visualize each (height, width) thing in a grid of size approx. sqrt(n) by sqrt(n)"""

        print "Data Shape : ", data.shape

        # normalize data for display
        data = (data - data.min()) / (data.max() - data.min())

        # force the number of filters to be square
        n = int(np.ceil(np.sqrt(data.shape[0])))
        padding = (((0, n ** 2 - data.shape[0]),
                    (0, 1), (0, 1))  # add some space between filters
                   + ((0, 0),) * (data.ndim - 3))  # don't pad the last dimension (if there is one)
        data = np.pad(data, padding, mode='constant', constant_values=0)  # pad with ones (white)

        # tile the filters into an image
        data = data.reshape((n, n) + data.shape[1:]).transpose((0, 2, 1, 3) + tuple(range(4, data.ndim + 1)))
        data = data.reshape((n * data.shape[1], n * data.shape[3]) + data.shape[4:])

        # show at display
        #print 'Data shape : ', data.shape , len(data.shape)
        img = 255 * data
        img = cv2.resize(img, (512, 512))
        img = np.array(img, dtype='uint8')
        img_c = cv2.applyColorMap(img, cv2.COLORMAP_JET)
        height, width, byteValue = img_c.shape
        byteValue = byteValue * width
        self.image = QtGui.QImage(img_c.data, width, height, byteValue, QtGui.QImage.Format_RGB888)
        self.ui.labelDisplay.setPixmap(QtGui.QPixmap(self.image))

def show_cam_on_image(img, mask):
    heatmap = cv2.applyColorMap(np.uint8(255*mask), cv2.COLORMAP_JET)
    heatmap = np.float32(heatmap) / 255
    cam = heatmap + np.float32(img)
    cam = cam / np.max(cam)
    cv2.imwrite("cam.jpg", np.uint8(255 * cam))

def grad_cam(input_model, image, category_index, layer_name):
    model = Sequential()
    model.add(input_model)

    nb_classes = 1000
    target_layer = lambda x: target_category_loss(x, category_index, nb_classes)
    model.add(Lambda(target_layer,
                     output_shape = target_category_loss_output_shape))

    loss = K.sum(model.layers[-1].output)
    conv_output =  [l for l in model.layers[0].layers if l.name is layer_name][0].output
    grads = normalize(K.gradients(loss, conv_output)[0])
    gradient_function = K.function([model.layers[0].input], [conv_output, grads])

    output, grads_val = gradient_function([image])
    output, grads_val = output[0, :], grads_val[0, :, :, :]

    weights = np.mean(grads_val, axis = (0, 1))
    cam = np.ones(output.shape[0 : 2], dtype = np.float32)

    for i, w in enumerate(weights):
        cam += w * output[:, :, i]

    cam = cv2.resize(cam, (224, 224))
    cam = np.maximum(cam, 0)
    heatmap = cam / np.max(cam)

    #Return to BGR [0..255] from the preprocessed image
    image = image[0, :]
    image -= np.min(image)
    image = np.minimum(image, 255)

    cam = cv2.applyColorMap(np.uint8(255*heatmap), cv2.COLORMAP_JET)
    cam = np.float32(cam) + np.float32(image)
    cam = 255 * cam / np.max(cam)
    return np.uint8(cam), heatmap

def save_cam_image(img, mask, filename):
    img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
    heatmap = cv2.applyColorMap(np.uint8(255.*mask), cv2.COLORMAP_JET)
    res = np.concatenate((img, heatmap), axis=1)
    cv2.imwrite(filename, res)

def generate_heatmap(self,section):

        if self.cache is None:
            logger.warning('Surface cache is not build yet.')
            return


        x,y = self.real_world_size['x'],self.real_world_size['y']
        x = max(1,int(x))
        y = max(1,int(y))

        filter_size = int(int(self.heatmap_detail * x)/2)*2 +1
        std_dev = int(filter_size /6.)
        self.heatmap = np.ones((y,x,4),dtype=np.uint8)
        all_gaze = []

        for frame_idx,c_e in enumerate(self.cache[section]):
            if c_e:
                frame_idx+=section.start
                for gp in self.gaze_on_srf_by_frame_idx(frame_idx,c_e['m_from_screen']):
                    if gp['confidence']>=self.g_pool.min_data_confidence:
                        all_gaze.append(gp['norm_pos'])

        if not all_gaze:
            logger.warning("No gaze data on surface for heatmap found.")
            all_gaze.append((-1.,-1.))
        all_gaze = np.array(all_gaze)
        all_gaze *= [self.real_world_size['x'],self.real_world_size['y']]
        hist,xedge,yedge = np.histogram2d(all_gaze[:,0], all_gaze[:,1],
                                            bins=[x,y],
                                            range=[[0, self.real_world_size['x']], [0,self.real_world_size['y']]],
                                            normed=False,
                                            weights=None)


        hist = np.rot90(hist)

        #smoothing..
        hist = cv2.GaussianBlur(hist,(filter_size,filter_size),std_dev)
        maxval = np.amax(hist)
        if maxval:
            scale = 255./maxval
        else:
            scale = 0

        hist = np.uint8( hist*(scale) )

        #colormapping
        c_map = cv2.applyColorMap(hist, cv2.COLORMAP_JET)

        self.heatmap[:,:,:3] = c_map
        self.heatmap[:,:,3] = 125
        self.heatmap_texture = Named_Texture()
        self.heatmap_texture.update_from_ndarray(self.heatmap)