In version 1.28.0, SNPE announced support for the TF MobileNet-V1-FPN-SSD model (https://developer.qualcomm.com/docs/snpe/revision_history.html). However I'm running into two errors during the TF_frozen_graph.pb -> DLC conversion:
RuntimeWarning: error_code=1004; error_message=Layer parameters combination is invalid. Layer concat: input WeightSharedConvolutionalBoxPredictor_1/ClassPredictor/BiasAdd:0 has size 40 along axis 2, should match output dim (80); error_component=Model Validation; line_no=926; thread_id=140525780473664...RuntimeWarning: error_code=802; error_message=Layer parameter value is invalid in GPU. Layer Postprocessor/Reshape_2:0 : input width = 51150 exceeds maximum image width 16384 for Adreno A530; error_component=GPU Runtime; line_no=647; thread_id=140525780473664
I'm using the default MobileNet-V1-FPN-SSDconfiguration from the TF Object Detection API: https://github.com/tensorflow/models/blob/master/research/object_detection/samples/configs/ssd_mobilenet_v1_fpn_shared_box_predictor_640x640_coco14_sync.config
Has anyone come accross the same issue?
SNPE: 1.31.0
Tensorflow: 1.14