def __init__(self, input_size=(640, 480)): self.C = config.Config() self.C.offset = True self.C.size_test = input_size input_shape_img = (self.C.size_test[0], self.C.size_test[1], 3) img_input = Input(shape=input_shape_img) # define the network prediction preds = nn.nn_p3p4p5(img_input, offset=self.C.offset, num_scale=self.C.num_scale, trainable=True) self.model = Model(img_input, preds) self.model_path = '/root/webapp/detector/models_weight' self.detec_hum(np.random.rand(200, 300, 3))
C.size_test = (1024, 2048) input_shape_img = (C.size_test[0], C.size_test[1], 3) img_input = Input(shape=input_shape_img) # define the base network (resnet here, can be MobileNet, etc) if C.network == 'resnet50': from keras_csp import resnet50 as nn elif C.network == 'mobilenet': from keras_csp import mobilenet as nn else: raise NotImplementedError('Not support network: {}'.format(C.network)) # define the network prediction preds = nn.nn_p3p4p5(img_input, offset=C.offset, num_scale=C.num_scale, trainable=True) model = Model(img_input, preds) if C.offset: w_path = 'output/valmodels/city_valMR/{}/off{}'.format(C.scale, exp_name) out_path = 'output/valresults/city_valMR/{}/off{}'.format( C.scale, exp_name) if not os.path.exists(out_path): os.makedirs(out_path) weight1 = os.path.join(w_path, 'best_val.hdf5') res_path = os.path.join(out_path, 'best_val') if not os.path.exists(res_path): os.makedirs(res_path) print(res_path)