From 6e08494a8883a8e304ef02ea352a11776ab2294c Mon Sep 17 00:00:00 2001 From: Nick White Date: Thu, 24 Jan 2019 17:56:14 +0000 Subject: Merge bucket-lines-{prob,hocr} into one tool called bucket-lines, that uses the filename extension to determine how to process the lines --- bucket-lines-hocr/bucket-lines-hocr.go | 91 ---------------------------- bucket-lines-prob/bucket-lines-prob.go | 56 ------------------ bucket-lines/bucket-lines.go | 105 +++++++++++++++++++++++++++++++++ parse/hocr/hocr.go | 9 +-- 4 files changed, 110 insertions(+), 151 deletions(-) delete mode 100644 bucket-lines-hocr/bucket-lines-hocr.go delete mode 100644 bucket-lines-prob/bucket-lines-prob.go create mode 100644 bucket-lines/bucket-lines.go diff --git a/bucket-lines-hocr/bucket-lines-hocr.go b/bucket-lines-hocr/bucket-lines-hocr.go deleted file mode 100644 index b35c824..0000000 --- a/bucket-lines-hocr/bucket-lines-hocr.go +++ /dev/null @@ -1,91 +0,0 @@ -package main - -// TODO: merge with -prob, using filename extension to determine what to do for each file - -import ( - "flag" - "fmt" - "image/png" - "io/ioutil" - "log" - "os" - "path/filepath" - "strings" - - "git.rescribe.xyz/testingtools/parse" - "git.rescribe.xyz/testingtools/parse/hocr" -) - -func detailsFromFile(f string) (parse.LineDetails, error) { - var newlines parse.LineDetails - - file, err := ioutil.ReadFile(f) - if err != nil { - return newlines, err - } - - h, err := hocr.Parse(file) - if err != nil { - return newlines, err - } - - pngfn := strings.Replace(f, ".hocr", ".png", 1) - pngf, err := os.Open(pngfn) - if err != nil { - return newlines, err - } - defer pngf.Close() - img, err := png.Decode(pngf) - if err != nil { - return newlines, err - } - - n := strings.Replace(filepath.Base(f), ".hocr", "", 1) - return hocr.GetLineDetails(h, img, n) -} - -func main() { - b := parse.BucketSpecs{ - // minimum confidence, name - { 0, "bad" }, - { 0.95, "95to98" }, - { 0.98, "98plus" }, - } - - flag.Usage = func() { - fmt.Fprintf(os.Stderr, "Usage: bucket-lines-hocr [-d dir] hocr1 [hocr2] [...]\n") - fmt.Fprintf(os.Stderr, "Copies image-text line pairs into different directories according\n") - fmt.Fprintf(os.Stderr, "to the average character probability for the line.\n") - fmt.Fprintf(os.Stderr, "This uses the x_wconf data in .hocr files, which it assumes will be.\n") - fmt.Fprintf(os.Stderr, "in the same directory as the line's image and text files. It can\n") - fmt.Fprintf(os.Stderr, "handle hocr where each character is tagged separately and hocr where\n") - fmt.Fprintf(os.Stderr, "only whole words are tagged.\n") - flag.PrintDefaults() - } - dir := flag.String("d", "buckets", "Directory to store the buckets") - flag.Parse() - if flag.NArg() < 1 { - flag.Usage() - os.Exit(1) - } - - lines := make(parse.LineDetails, 0) - - for _, f := range flag.Args() { - newlines, err := detailsFromFile(f) - if err != nil { - log.Fatal(err) - } - - for _, l := range newlines { - lines = append(lines, l) - } - } - - stats, err := parse.BucketUp(lines, b, *dir) - if err != nil { - log.Fatal(err) - } - - parse.PrintBucketStats(os.Stdout, stats) -} diff --git a/bucket-lines-prob/bucket-lines-prob.go b/bucket-lines-prob/bucket-lines-prob.go deleted file mode 100644 index 728268d..0000000 --- a/bucket-lines-prob/bucket-lines-prob.go +++ /dev/null @@ -1,56 +0,0 @@ -package main - -import ( - "flag" - "fmt" - "log" - "os" - - "git.rescribe.xyz/testingtools/parse" - "git.rescribe.xyz/testingtools/parse/prob" -) - -func main() { - b := parse.BucketSpecs{ - // minimum confidence, name - { 0, "bad" }, - { 0.95, "95to98" }, - { 0.98, "98plus" }, - } - - flag.Usage = func() { - fmt.Fprintf(os.Stderr, "Usage: bucket-lines-prob [-d dir] prob1 [prob2] [...]\n") - fmt.Fprintf(os.Stderr, "Copies image-text line pairs into different directories according\n") - fmt.Fprintf(os.Stderr, "to the average character probability for the line.\n") - fmt.Fprintf(os.Stderr, "This uses the .prob files generated by ocropy-rpred's --probabilities\n") - fmt.Fprintf(os.Stderr, "option, which it assumes will be in the same directory as the line's\n") - fmt.Fprintf(os.Stderr, "image and text files.\n") - flag.PrintDefaults() - } - dir := flag.String("d", "buckets", "Directory to store the buckets") - flag.Parse() - if flag.NArg() < 1 { - flag.Usage() - os.Exit(1) - } - - lines := make(parse.LineDetails, 0) - - for _, f := range flag.Args() { - newlines, err := prob.GetLineDetails(f) - if err != nil { - log.Fatal(err) - } - - for _, l := range newlines { - lines = append(lines, l) - } - } - - stats, err := parse.BucketUp(lines, b, *dir) - if err != nil { - log.Fatal(err) - } - - parse.PrintBucketStats(os.Stdout, stats) -} diff --git a/bucket-lines/bucket-lines.go b/bucket-lines/bucket-lines.go new file mode 100644 index 0000000..ad73fcd --- /dev/null +++ b/bucket-lines/bucket-lines.go @@ -0,0 +1,105 @@ +package main + +import ( + "flag" + "fmt" + "image/png" + "io/ioutil" + "log" + "os" + "path/filepath" + "strings" + + "git.rescribe.xyz/testingtools/parse" + "git.rescribe.xyz/testingtools/parse/hocr" + "git.rescribe.xyz/testingtools/parse/prob" +) + +// TODO: maybe move this into hocr.go +func detailsFromHocr(f string) (parse.LineDetails, error) { + var newlines parse.LineDetails + + file, err := ioutil.ReadFile(f) + if err != nil { + return newlines, err + } + + h, err := hocr.Parse(file) + if err != nil { + return newlines, err + } + + pngfn := strings.Replace(f, ".hocr", ".png", 1) + pngf, err := os.Open(pngfn) + if err != nil { + return newlines, err + } + defer pngf.Close() + img, err := png.Decode(pngf) + if err != nil { + return newlines, err + } + + n := strings.Replace(filepath.Base(f), ".hocr", "", 1) + return hocr.GetLineDetails(h, img, n) +} + +func main() { + // TODO: Allow different specs to be used for .prob vs .hocr. Do this + // by adding a field to LineDetails that is linked to a named + // BucketSpecs. + b := parse.BucketSpecs{ + // minimum confidence, name + { 0, "bad" }, + { 0.95, "95to98" }, + { 0.98, "98plus" }, + } + + flag.Usage = func() { + fmt.Fprintf(os.Stderr, "Usage: bucket-lines [-d dir] [hocr1] [prob1] [hocr2] [...]\n") + fmt.Fprintf(os.Stderr, "Copies image-text line pairs into different directories according\n") + fmt.Fprintf(os.Stderr, "to the average character probability for the line.\n\n") + fmt.Fprintf(os.Stderr, "Both .hocr and .prob files can be processed.\n\n") + fmt.Fprintf(os.Stderr, "For .hocr files, the x_wconf data is used to calculate confidence.\n\n") + fmt.Fprintf(os.Stderr, "The .prob files are generated using ocropy-rpred's --probabilities\n") + fmt.Fprintf(os.Stderr, "option.\n\n") + fmt.Fprintf(os.Stderr, "The .prob and .hocr files are assumed to be in the same directory\n") + fmt.Fprintf(os.Stderr, "as the line's image and text files.\n\n") + flag.PrintDefaults() + } + dir := flag.String("d", "buckets", "Directory to store the buckets") + flag.Parse() + if flag.NArg() < 1 { + flag.Usage() + os.Exit(1) + } + + var err error + lines := make(parse.LineDetails, 0) + + for _, f := range flag.Args() { + var newlines parse.LineDetails + switch ext := filepath.Ext(f); ext { + case ".prob": + newlines, err = prob.GetLineDetails(f) + case ".hocr": + newlines, err = detailsFromHocr(f) + default: + log.Printf("Skipping file '%s' as it isn't a .prob or .hocr\n", f) + } + if err != nil { + log.Fatal(err) + } + + for _, l := range newlines { + lines = append(lines, l) + } + } + + stats, err := parse.BucketUp(lines, b, *dir) + if err != nil { + log.Fatal(err) + } + + parse.PrintBucketStats(os.Stdout, stats) +} diff --git a/parse/hocr/hocr.go b/parse/hocr/hocr.go index f7cac05..c03b73a 100644 --- a/parse/hocr/hocr.go +++ b/parse/hocr/hocr.go @@ -1,9 +1,10 @@ package hocr -// TODO: consider making GetLineDetails() a function of Hocr, so could do a -// similar thing with prob format files too, and then fire them both -// off a generic interface, potentially. -// TODO: Parse line name to zero pad line numbers, so they come out in the correct order +// TODO: Parse line name to zero pad line numbers, so they can +// be sorted easily +// TODO: have same filename format as .prob uses, so include base +// dirname, and don't include line numbers if there's only +// one line in the hocr import ( "encoding/xml" -- cgit v1.2.1-24-ge1ad