Skip to content

allow varying size inputs, with varying size channels #31

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
fdan wants to merge 1 commit into
base: master
Choose a base branch
from
Open

allow varying size inputs, with varying size channels #31

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 13 additions & 11 deletions Plugins/Client/MLClient.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -406,6 +406,8 @@ bool MLClient::processImage(const std::string& hostStr, int port,

// Parse image. TODO: Check for multiple inputs, different channel size
for (int i = 0; i < node_inputs(); i++) {
const ChannelSet readChannels = input(i)->info().channels();

// Create an ImagePlane, and read each input into it.
// Get our input & sanity check
DD::Image::Iop* inputIop = dynamic_cast<DD::Image::Iop*>(input(i));
Expand All @@ -429,17 +431,16 @@ bool MLClient::processImage(const std::string& hostStr, int port,

// Set our input bounding box, this is what our inputs can give us.
Box imageBounds = inputIop->info();
// We're going to clip it to our format.
imageBounds.intersect(imageFormat);

const int fx = imageBounds.x();
const int fy = imageBounds.y();
const int fr = imageBounds.r();
const int ft = imageBounds.t();

// Request our default channels, for our own bounding box
inputIop->request(fx, fy, fr, ft, kDefaultChannels, 0);
inputIop->request(fx, fy, fr, ft, readChannels, 0);
// Let's assume everything went fine, and fetch our plane
ImagePlane plane(imageBounds, /*packed*/ true, kDefaultChannels, kDefaultNumberOfChannels);
ImagePlane plane(imageBounds, /*packed*/ true, readChannels, readChannels.size());
inputIop->fetchPlane(plane);

// Sanity check that that the plane was filled successfully, and nothing
Expand All @@ -457,12 +458,13 @@ bool MLClient::processImage(const std::string& hostStr, int port,

// Set up our message
mlserver::Image* image = requestInference->add_images();
image->set_width(imageFormat.w());
image->set_height(imageFormat.h());
image->set_channels(kDefaultNumberOfChannels);
image->set_width(fr);
image->set_height(ft);
image->set_channels(readChannels.size());
// image->set_channels(kDefaultNumberOfChannels);

// Set up our temp contiguous buffer
size_t byteBufferSize = imageFormat.w() * imageFormat.h() * kDefaultNumberOfChannels * sizeof(float);
size_t byteBufferSize = fr * ft * readChannels.size() * sizeof(float);
if (byteBufferSize == 0) {
errorMsg = "Image size is zero.";
return false;
Expand All @@ -476,11 +478,11 @@ bool MLClient::processImage(const std::string& hostStr, int port,
// can't guarantee that it's contiguous, or packed in the
// expected way.
float* floatBuffer = (float*)byteBuffer;
for (int z = 0; z < kDefaultNumberOfChannels; z++) {
const int chanStride = z * imageFormat.w() * imageFormat.h();
for (int z = 0; z < readChannels.size(); z++) {
const int chanStride = z * fr * ft;

for (int ry = fy; ry < ft; ry++) {
const int rowStride = ry * imageFormat.w();
const int rowStride = ry * fr;

ImageTileReadOnlyPtr tile = plane.readableAt(ry, z);
for (int rx = fx, currentPos = 0; rx < fr; rx++) {
Expand Down