Merge branch 'master' of https://www.github.com/thedanielsun/ThreatEx…

…change
facebook · Aug 1, 2019 · 272084c · 272084c
2 parents ef45f6e + 0062694
commit 272084c
Show file tree

Hide file tree

Showing 8 changed files with 22 additions and 5 deletions.
diff --git a/hashing/README.md b/hashing/README.md
@@ -1,6 +1,6 @@
 Please see:
 
-* [Facebook Newsroom post on the open-source release of TMK+PDQF and PDQ](https://newsroom.fb.com/news/2019/08/link-name-tbd])
+* [Facebook Newsroom post on the open-source release of TMK+PDQF and PDQ](https://newsroom.fb.com/news/2019/08/open-source-photo-video-matching])
 * [Detailed information on algorithm sketch, details, and evaluation](https://github.com/facebook/ThreatExchange/blob/master/hashing/hashing.pdf)
 * [Implementation and setup information about TMK+PDQF](https://github.com/facebook/ThreatExchange/blob/master/hashing/tmk/README.md)
 * [Implementation and setup information about PDQ](https://github.com/facebook/ThreatExchange/blob/master/hashing/pdq/README.md)
diff --git a/hashing/pdq/README.md b/hashing/pdq/README.md
@@ -10,7 +10,7 @@ as well as
 [../hashing.pdf](https://github.com/facebook/ThreatExchange/blob/master/hashing/hashing.pdf)
 within this repository.
 
-See also https://newsroom.fb.com/news/2019/08/to-do-add-this-link-once-published for context.
+See also https://newsroom.fb.com/news/2019/08/open-source-photo-video-matching for context.
 
 As of November 2018 there are C++, PHP, and Java implementations.  Details are in the `*/README.md` files.
 

diff --git a/hashing/pdq/cpp/Makefile b/hashing/pdq/cpp/Makefile
@@ -10,6 +10,8 @@ CCDBG=$(CC) -g $(CFLAGS) $(IFLAGS) $(WFLAGS)
 CCSAN=clang -fsanitize=address -g $(CFLAGS) $(IFLAGS) $(WFLAGS)
 
 # ================================================================
+# Please replace CImg.h with your own system image-decoder and edit
+# io/pdqio.cpp.
 LIBHDRS=\
  common/pdqbasetypes.h \
  common/pdqhashtypes.h \
@@ -19,7 +21,8 @@ LIBHDRS=\
  downscaling/downscaling.h \
  io/pdqio.h \
  index/mih.h \
- hashing/torben.h
+ hashing/torben.h \
+ ./CImg.h
 
 LIBSRCS=\
  common/pdqhashtypes.cpp \
@@ -73,11 +76,12 @@ test-hamming: bin/test-hamming.cpp $(LIBSRCS) $(LIBHDRS)
 
 pdq-photo-hasher: bin/pdq-photo-hasher.cpp $(LIBSRCS) $(LIBHDRS)
 	$(CCOPT) bin/pdq-photo-hasher.cpp $(LIBSRCS) -o pdq-photo-hasher $(LFLAGS)
+
 # Profile version. Usage:
 # * make pdq-photo-hasher-prof
 # * pdq-photo-hasher-prof {arguments>
 # * gprof pdq-photo-hasher-prof gmon.out > myfile.txt
-# Note: works on Linux; not on OSX.
+# Note: this profiler command works on Linux; not on OSX.
 pdq-photo-hasher-prof: bin/pdq-photo-hasher.cpp $(LIBSRCS) $(LIBHDRS)
 	$(CCDBG) -pg bin/pdq-photo-hasher.cpp $(LIBSRCS) -o pdq-photo-hasher-prof $(LFLAGS) -lpthread
 

diff --git a/hashing/pdq/cpp/README.md b/hashing/pdq/cpp/README.md
@@ -6,6 +6,7 @@ Please see `../README.md` and `../../README.md` for context, and `../../hashing.
 
 * C++ 11 or higher
 * `CImg.h` is included for reference, although note that its license does not allow commercial use. It is expected that your company will already have image-processing logic. Dependencies of this code on `CImg.h` are confined solely to `io/pdqio.h` and `io/pdqio.cpp` which you can customize for your site.
+* ImageMagick or other JPEG/PNG libraries: for example, `brew install imagemagick` on MacOSX
 
 # Contact
 

diff --git a/hashing/pdq/cpp/bin/pdq-photo-hasher.cpp b/hashing/pdq/cpp/bin/pdq-photo-hasher.cpp
@@ -43,6 +43,7 @@ static void usage(char* argv0, int rc) {
       fp,
       "-k: Continue to next image after image errors, but still exit "
       "1 afterward.\n");
+  fprintf(fp, "--info: show information about the image-decoder library.\n");
   exit(rc);
 }
 
@@ -69,6 +70,10 @@ int main(int argc, char* argv[]) {
     } else if (!strcmp(argv[argi], "--help")) {
       usage(argv[0], 0);
 
+    } else if (!strcmp(argv[argi], "--info")) {
+      showDecoderInfo();
+      exit(0);
+
     } else if (
         !strcmp(argv[argi], "-i") || !strcmp(argv[argi], "--files-on-stdin")) {
       files_on_stdin = true;

diff --git a/hashing/pdq/cpp/io/pdqio.cpp b/hashing/pdq/cpp/io/pdqio.cpp
@@ -18,6 +18,11 @@ namespace hashing {
 // so we use off-the-shelf downsampling to get to an intermediate size.
 const int DOWNSAMPLE_DIMS = 512;
 
+// ----------------------------------------------------------------
+void showDecoderInfo() {
+  cimg_library::cimg::info();
+}
+
 // ----------------------------------------------------------------
 // Returns matrix as num_rows x num_cols in row-major order.
 // The caller must free the return value.

diff --git a/hashing/pdq/cpp/io/pdqio.h b/hashing/pdq/cpp/io/pdqio.h
@@ -24,6 +24,8 @@ float* loadFloatLumaFromCImg(
   int& numCols
 );
 
+void showDecoderInfo();
+
 bool pdqHash256FromFile(
   const char* filename,
   Hash256& hash,

diff --git a/hashing/tmk/README.md b/hashing/tmk/README.md
@@ -5,7 +5,7 @@ produced in conjunction with Facebook AI Research ('FAIR').
 
 * Full details are located in the [hashing.pdf](https://github.com/facebook/ThreatExchange/blob/master/hashing/hashing.pdf) document.
 
-* See https://newsroom.fb.com/news/2019/08/to-do-add-this-link-once-published for context.
+* See https://newsroom.fb.com/news/2019/08/open-source-photo-video-matching for context.
 
 * Researchers in video copy-detection algorithms may also be interested in
   https://github.com/facebookresearch/videoalignment for a research version of