Closed f0rodo closed 2 years ago
The above ML pipeline isn't correct. I'm attaching the correct Runefile and rune below.
version: 1
image: runicos/base
pipeline:
image:
capability: IMAGE
outputs:
- type: u8
dimensions: [1,513,513,3]
args:
height: 513
width: 513
pixel-format: "@PixelFormat::RGB"
deep_lab:
model: "./food_segmenter.tflite"
inputs:
- image
outputs:
- type: f32
dimensions: [1,513,513,26]
segmentation_map:
proc-block: "hotg-ai/proc-blocks@v0.11.0#segment_output"
inputs:
- deep_lab
outputs:
- type: u32
dimensions: [513, 513]
- type: u32
dimensions: [5]
label:
proc-block: "hotg-ai/proc-blocks@v0.11.0#label"
inputs:
- segmentation_map.1
outputs:
- type: Utf8
dimensions: [3]
args:
wordlist: |
background
vegetables | leafy_greens
vegetables | stem_vegetables
vegetables | non-starchy_roots
vegetables | other
fruits
protein | meat
protein | poultry
protein | seafood
protein | eggs
protein | beans/nuts
starches/grains | baked_goods
starches/grains | rice/grains/cereals
starches/grains | noodles/pasta
starches/grains | starchy_vegetables
starches/grains | other
soups/stews
herbs/spices
dairy
snacks
sweets/desserts
beverages
fats/oils/sauces
food_containers
dining_tools
other_food
serial:
out: SERIAL
inputs:
- segmentation_map.0
- label
The Runic app is crashing because we haven't added the functionality to allow a user to run rune which requires uploading a label file.
I am trying to build a segmenter app and run it on runic app.
https://tfhub.dev/google/seefood/segmenter/mobile_food_segmenter_V1/1