SIMS
SIMS copied to clipboard
xjqicuhk
Error when run order.py
I was perform the inference on my own semantic masks with resolution "1024 * 2048". According the "Inference" part says, first run test_connect_component_mask.m、search_segments.m、test_generate_transform.m、transformation.py and test_generate_order.m, it's all good. But when run the order.py i got this error:
Traceback (most recent call last): File "/home/x00478666/pycharm-community-2018.2/helpers/pydev/pydevd.py", line 1664, in <module> main() File "/home/x00478666/pycharm-community-2018.2/helpers/pydev/pydevd.py", line 1658, in main globals = debugger.run(setup['file'], None, None, is_module) File "/home/x00478666/pycharm-community-2018.2/helpers/pydev/pydevd.py", line 1068, in run pydev_imports.execfile(file, globals, locals) # execute the script File "/home/x00478666/pycharm-community-2018.2/helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile exec(compile(contents+"\n", file, 'exec'), glob, loc) File "/home/distributed-training/x00478666/SIMS/SIMS-master/python_code/order.py", line 259, in <module> sess.run(proposal_score, feed_dict={data: tmp_data})) ValueError: could not broadcast input array from shape (4,5,20) into shape (1,20)
and in order.py i set this:
batch_size = 1 crop_size_h = 1024 crop_size_w = 2048 input_dim = 6
Did i miss something? Best regards.
It's caused by the input "data: tmp_data" cannot fit the parameters defined in the place holder for variable "proposal_score"
On Wed, Mar 27, 2019 at 6:00 PM CGod1991 [email protected] wrote:
I was perform the inference on my own semantic masks with resolution "1024
- 2048". According the "Inference" part says, first run test_connect_component_mask.m、search_segments.m、test_generate_transform.m、transformation.py and test_generate_order.m, it's all good. But when run the order.py i got this error: Traceback (most recent call last): File "/home/x00478666/pycharm-community-2018.2/helpers/pydev/pydevd.py", line 1664, in
main() File "/home/x00478666/pycharm-community-2018.2/helpers/pydev/pydevd.py", line 1658, in main globals = debugger.run(setup['file'], None, None, is_module) File "/home/x00478666/pycharm-community-2018.2/helpers/pydev/pydevd.py", line 1068, in run pydev_imports.execfile(file, globals, locals) # execute the script File "/home/x00478666/pycharm-community-2018.2/helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile exec(compile(contents+"\n", file, 'exec'), glob, loc) File "/home/distributed-training/x00478666/SIMS/SIMS-master/python_code/order.py", line 259, in sess.run(proposal_score, feed_dict={data: tmp_data})) ValueError: could not broadcast input array from shape (4,5,20) into shape (1,20) and this is my order.py content: `from future import division
import cv2 import os import scipy.io import time
import numpy as np import tensorflow as tf import tensorflow.contrib.slim as slim
os.environ['CUDA_VISIBLE_DEVICES'] = '6'
training_phase = False if (training_phase): label_path = "../traindata/label_refine/" mat_path = "../traindata/order/" save_model_path = "../trainedmodels/order" else: label_path = "../testdata/video/label_refine/" mat_path = '../testdata/video/order/' save_result_path = "../result/order/data/" save_model_path = "../trainedmodels/order"
os.system('nvidia-smi -q -d Memory |grep -A4 GPU|grep Free >tmp') os.environ['CUDA_VISIBLE_DEVICES'] = str(np.argmax([int(x.split()[2]) for x in open('tmp', 'r').readlines()])) os.system('rm tmp')
batch_size = 1 crop_size_h = 1024 crop_size_w = 2048 input_dim = 6 sp = 1024 topk = 10
sess = tf.Session()
def lrelu(x): return tf.maximum(0.2 * x, x)
MEAN_VALUES = np.array([123.6800, 116.7790, 103.9390]).reshape((1, 1, 1, 3))
def build_net(ntype, nin, nwb=None, name=None): if ntype == 'conv': return tf.nn.relu(tf.nn.conv2d(nin, nwb[0], strides=[1, 1, 1, 1], padding='SAME', name=name) + nwb[1]) elif ntype == 'pool': return tf.nn.avg_pool(nin, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
def get_weight_bias(vgg_layers, i): weights = vgg_layers[i][0][0][2][0][0] weights = tf.constant(weights) bias = vgg_layers[i][0][0][2][0][1] bias = tf.constant(np.reshape(bias, (bias.size))) return weights, bias
def one_hot(label):
print(label.shape)
output = np.zeros((1, 100, label.shape[0], label.shape[1]), dtype=np.float32); for i in range(label.shape[0]): for j in range(label.shape[1]): if label[i, j] > 0: output[0, label[i, j] - 1, i, j] = 1 return output
def proposal_classifier(data): conv1_selection = slim.repeat(data, 2, slim.conv2d, 64, [3, 3], rate=1, normalizer_fn=slim.layer_norm, activation_fn=lrelu, scope='g_selection_conv1') pool1_selection = slim.avg_pool2d(conv1_selection, [3, 3], stride=2, padding='SAME', scope='g_selection_pool1')
conv2_selection = slim.repeat(pool1_selection, 2, slim.conv2d, 128, [3, 3], rate=1, normalizer_fn=slim.layer_norm,
activation_fn=lrelu, scope='g_selection_conv2')pool2_selection = slim.max_pool2d(conv2_selection, [3, 3], stride=2, padding='SAME', scope='g_selection_pool2')
conv3_selection = slim.repeat(pool2_selection, 3, slim.conv2d, 256, [3, 3], rate=1, normalizer_fn=slim.layer_norm,
activation_fn=lrelu, scope='g_selection_conv3')pool3_selection = slim.max_pool2d(conv3_selection, [3, 3], stride=2, padding='SAME', scope='g_selection_pool3')
conv4_selection = slim.repeat(pool3_selection, 3, slim.conv2d, 512, [3, 3], rate=1, normalizer_fn=slim.layer_norm,
activation_fn=lrelu, scope='g_selection_conv4')pool4_selection = slim.max_pool2d(conv4_selection, [3, 3], stride=2, padding='SAME', scope='g_selection_pool4')
conv5_selection = slim.repeat(pool4_selection, 3, slim.conv2d, 512, [3, 3], rate=1, normalizer_fn=slim.layer_norm,
activation_fn=lrelu, scope='g_selection_conv5')pool5_selection = slim.max_pool2d(conv5_selection, [3, 3], stride=2, padding='SAME', scope='g_selection_pool5')
pool6_selection = slim.avg_pool2d(pool5_selection, [8, 12], stride=[8, 12], padding='VALID',
scope='g_selection_pool6')fc_selection = slim.conv2d(pool6_selection, 1024, [1, 1], rate=1, normalizer_fn=slim.layer_norm,
activation_fn=lrelu, scope='g_selection_conv6')fc_selection = slim.dropout(fc_selection, 0.5, is_training=training_phase, scope='drop7_selection_norm')
score = slim.conv2d(fc_selection, 20, [1, 1], rate=1,
activation_fn=None, scope='g_selection_classify')score = tf.squeeze(score)
return score
with tf.variable_scope(tf.get_variable_scope()) as scope: label = tf.placeholder(tf.int32, [batch_size, 1]) data = tf.placeholder(tf.float32, [batch_size, None, None, input_dim])
sp = 256
proposal_score = proposal_classifier(data)
weight = tf.placeholder(tf.float32)
label = tf.squeeze(label)
G_loss = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=label, logits=proposal_score)
G_loss = tf.reduce_mean(G_loss)
lr = tf.placeholder(tf.float32)
t_vars = tf.trainable_variables() for var in t_vars: if var.name.startswith('g_'): print(var.name)
G_opt = tf.train.AdamOptimizer(learning_rate=lr).minimize(loss=G_loss, var_list=[var for var in t_vars if var.name.startswith('g_')])
saver = tf.train.Saver(max_to_keep=1000) sess.run(tf.global_variables_initializer()) ckpt = tf.train.get_checkpoint_state(save_model_path + "/0100") if ckpt: print('loaded ' + ckpt.model_checkpoint_path) saver.restore(sess, ckpt.model_checkpoint_path)
best = np.inf
g_loss = np.zeros(3000, dtype=float)
tmp_label = np.zeros((batch_size, 1), dtype=np.int32) tmp_data = np.zeros((batch_size, crop_size_h, crop_size_w, input_dim), dtype=float) tmp_mask = np.zeros((batch_size, 1), dtype=float)
total_number = 2975 base_lr = 1e-6 if (training_phase): for epoch in range(1, 101): tmp_label = np.zeros((batch_size, 1), dtype=np.int32) tmp_data = np.zeros((batch_size, crop_size_h, crop_size_w, input_dim), dtype=float) tmp_mask = np.zeros((batch_size, 1), dtype=float) if os.path.isdir(save_model_path + "/%04d" % epoch): continue
tmp_list = np.random.permutation(total_number) + 1 global_ind = 0 while (global_ind < total_number): st = time.time() try: dic = scipy.io.loadmat(mat_path + "%08d.mat" % tmp_list[global_ind]) except: print("cannot load" + "%08d.mat" % tmp_list[global_ind]) global_ind = global_ind + 1 continue topk_proposal = dic['semantic_segment_mask'].astype(np.float32) if (topk_proposal.ndim == 0): global_ind = global_ind + 1 continue corr_label = dic['semantic_segment_label'].astype(np.int32) topk_proposal = topk_proposal / 255.0 semantic = cv2.imread(label_path + "/%08d.png" % tmp_list[global_ind]).astype(np.float32) semantic = semantic / 255.0 tmp_semantic = semantic.copy() tmp_semantic[:, :, 0] = semantic[:, :, 2].copy() tmp_semantic[:, :, 1] = semantic[:, :, 1].copy() tmp_semantic[:, :, 2] = semantic[:, :, 0].copy() if (topk_proposal.ndim == 3): tmp_data[0, :, :, 0:3] = topk_proposal tmp_data[0, :, :, 3:6] = tmp_semantic tmp_label[0, :] = corr_label if (topk_proposal.ndim == 4): num_proposal = topk_proposal.shape[3] if (num_proposal >= batch_size): tmp_data[0:batch_size, :, :, 0:3] = np.transpose(topk_proposal[:, :, :, 0:batch_size], [3, 0, 1, 2]) tmp_data[0:batch_size, :, :, 3:6] = np.tile(np.expand_dims(tmp_semantic, axis=0), (batch_size, 1, 1, 1)) tmp_label[0:batch_size, :] = corr_label[0:batch_size, :] else: tmp_data[0:num_proposal, :, :, 0:3] = np.transpose(topk_proposal[:, :, :, 0:num_proposal], [3, 0, 1, 2]) tmp_data[0:num_proposal, :, :, 3:6] = np.tile(np.expand_dims(tmp_semantic, axis=0), (num_proposal, 1, 1, 1)) tmp_label[0:num_proposal, :] = corr_label[0:num_proposal, :] print(tmp_label[0]) _, G_current, lr_value, proposal_score_val = sess.run([G_opt, G_loss, lr, proposal_score], feed_dict={label: np.squeeze(tmp_label), data: tmp_data, lr: min(base_lr * np.power(1.1, epoch - 1), 1e-5)}) g_loss[global_ind] = G_current global_ind = global_ind + 1 print("%d %d %.2f %.2f %.2f %.6f" % ( epoch, global_ind, G_current, np.mean(g_loss[np.where(g_loss)]), time.time() - st, lr_value)) if epoch % 1 == 0: os.makedirs(save_model_path + "/%04d" % epoch) target = open(save_model_path + "/%04d/score.txt" % epoch, 'w') target.write("%f" % np.mean(g_loss[np.where(g_loss)])) target.close() saver.save(sess, save_model_path + "/model.ckpt") if epoch % 20 == 0: saver.save(sess, save_model_path + "/%04d/model.ckpt" % epoch)else: if not os.path.isdir(save_result_path): os.makedirs(save_result_path)
ind = 0
for label_name in os.listdir(label_path):
ind += 1 print(ind) dic = scipy.io.loadmat(os.path.join(mat_path, label_name.replace(".png", ".mat"))) proposal_all = dic['semantic_segment_mask'].astype(np.float32) / 255.0 if (proposal_all.shape[0] == 0): continue if (proposal_all.ndim == 3): print('expanding....') proposal_all = np.expand_dims(proposal_all, axis=3) semantic = cv2.imread(os.path.join(label_path, label_name)) semantic = semantic.astype(np.float32) / 255.0 tmp_semantic = semantic.copy() tmp_semantic[:, :, 0] = semantic[:, :, 2].copy() tmp_semantic[:, :, 1] = semantic[:, :, 1].copy() tmp_semantic[:, :, 2] = semantic[:, :, 0].copy() final_pred = np.zeros((proposal_all.shape[3], 20), dtype=np.float32) count_all = proposal_all.shape[3] index = 0 while (count_all > 0): tmp_data = np.zeros((batch_size, crop_size_h, crop_size_w, input_dim), dtype=float) if (count_all > batch_size): tmp_data[0:batch_size, :, :, 0:3] = np.transpose(proposal_all[:, :, :, index:index + batch_size], [3, 0, 1, 2]) tmp_data[0:batch_size, :, :, 3:6] = np.tile(np.expand_dims(tmp_semantic, axis=0), (batch_size, 1, 1, 1)) final_pred[index:index + batch_size, :] = np.squeeze( sess.run(proposal_score, feed_dict={data: tmp_data})) count_all = count_all - batch_size index = index + batch_size else: tmp_data[0:count_all, :, :, 0:3] = np.transpose(proposal_all[:, :, :, index:index + count_all], [3, 0, 1, 2]) tmp_data[0:count_all, :, :, 3:6] = np.tile(np.expand_dims(tmp_semantic, axis=0), (count_all, 1, 1, 1)) final_pred[index:index + count_all, :] = np.squeeze( sess.run(proposal_score, feed_dict={data: tmp_data}))[0:count_all, :] index = index + count_all count_all = 0 scipy.io.savemat(os.path.join(save_result_path, label_name.replace(".png", ".mat")), {'prediction': final_pred})`
Did i miss something? Best regards.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/xjqicuhk/SIMS/issues/7, or mute the thread https://github.com/notifications/unsubscribe-auth/Ak1JBLRFA5v_axmgIh_8kEkk_bsh2MV0ks5va0FIgaJpZM4cNYi0 .
