System information (version)

• OpenCV => 4.3.0

Detailed description

Hi, I'm trying to generate the text graph representation for MobileNetV3-SSD from tensorflow to be used in DNN module. I fine-tuned the ssd_mobilenet_v3_small_coco_2020_01_14 on a custom class and everything was fine in the text graph generation, everything is also working with the DNN module.
Then I resumed training modifying the pipeline.config again with some data augmentation and dropout. When I genereate the text-graph representation I have no erros.. but then when I import with

net = cv.dnn_DetectionModel('frozen_inference_graph.pb', 'cvgraph.pbtxt')

this error appear:

"Input layer not found: FeatureExtractor/MobilenetV3/expanded_conv_8/expansion_output in function 'connect"

The same network, with different pipeline.config generate two different text graph representations, but the second one doesn't even work with DNN module. Strange thing is that if I use the second frozen_inference_graph with the first cvgraph.pbtxt everything works!

pipeline.zip

Steps to reproduce

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues,
  answers.opencv.org, Stack Overflow, etc and have not found solution
• I updated to latest OpenCV version and the issue is still there
• There is reproducer code and related data files: videos, images, onnx, etc